The Definitive Internet Resource for the Semiconductor Manufacturing Industry™

Industry Search

Korean Site
Chinese Site
Japanese Site

  
Advanced Search    Add or Edit a Listing

Semiconductor International June 1999

Darryl Weddington, John Bodnar
and Mo Jahanbani

Advanced Products Research
and Development Laboratory,
Motorola, Austin, Texas

Roger Blum,
Millipore Corp., Bedford, Mass.

An Assessment of
a Dilute
Hydrofluoric Acid Purifier

Wet cleans account for as much as a third of all process steps in VLSI manufacturing, depending on the process flow and tool configuration.1 A large number of these steps use dilute hydrofluoric (DHF) acid to remove silicon oxide, native or chemical oxide, metallic contamination and particles by slightly etching the silicon surface. DHF purity is critical due to the acid's tendency to deposit metal contaminants on the exposed silicon surface in the active area where the gate oxides are to be grown. The ions of noble metals (such as copper, zinc and iron) deposit on the silicon surface by elemental reduction, which is absorption of metal species with electrochemical reduction to metal on the bare silicon surface. These metallic deposits degrade the device's electrical properties by creating generation-recombination centers that affect minority carrier lifetime.2,3

Our facility currently does not monitor metal contamination on a real-time basis. TXRF (total x-ray reflection fluorescence) analysis is performed on a bare silicon p-type wafer weekly to monitor the cleanliness of the DHF bath. TXRF and ICP-MS (inductively coupled plasma mass spectrometry) data collected in DHF solutions do not indicate a challenge with metallics. The sources of metal contamination in typical processing (chemicals, equipment and incoming water) are usually low enough to allow for extended bath life. The utility of an acid purifier then is to control excursions when nonstandard events occur. This article discusses the filter's capability to remove metallics during such excursions.

Purifier tests

The investigation of the Millipore Chempure polyethylene purifier was performed in a DHF immersion bath. The HF solution was dilute and (0.5% by weight HF), using a concentration of 100 parts of ultrapure deionized (UPDI) water to one part 49% hydrofluoric acid. All chemicals are Gigabit grade (<100 ppt per element) and the UPDI water with combined metallic concentration ¾ 0.1 ppb. Increasing the purity of the incoming chemicals including UPDI has reduced metallic impurities present in the baths. Unfortunately, contamination at the point of use also increases as more wafers are processed in the bath.4 The DHF solution is chilled to 22°C.

Two evaluations were conducted to determine how effective the point of use chemical purifier filter was in a series of tests in a 100:1 HF recirculation bath. These tests were conducted over several days, focusing on ion removal from the DHF bath, as well as particle and etch performance. The DHF chemical was changed out each day except where noted on Tables 1a and 1b.

In the first experiment, the bath was spiked each day with 30 ppb of mixed ions. ICP-MS, TXRF, particle and etch performance were the analysis tools used to determine the effectiveness of the purifier. The process characterization (witness) wafers used during this evaluation were immersed for two minutes in the DHF solution. At the end of first experiment, the tank's concentration was converted from 100:1 to 10:1 to determine the effects of the lower pH solution on the purifier's membrane (see Table 1a for the testing schedule).
Table 1a Test 1 Matrix
Run
Day
Hour
Wafers Processed
TXRF
Sample
PC & etch
50 Wafer
1
1
0
0
X
X
X
 
2
1
spiked
50
X
X
X
 
3
1
1
100
X
X
X
 
4
1
4
800
X
X
X
X
5
1
24
4650
X
X
X
X
6
2
0
0
X
X
X
 
7
2
spiked
50
X
X
X
 
8
2
1
100
X
X
X
 
9
2
4
800
X
 
X
X
10
2
24
4650
X
X
X
X
11
3
0
0
X
X
X
X
12
3
spiked
50
X
X
X
 
13
3
1
100
X
X
X
 
14
3
4
800
X
X
X
X
15
3
72
4650
X
X
X
X
16
4
0
0
X
X
X
 
17
5
0
0
X
X
X
 
 
Table 1b Test 2 Matrix
Run
Day
Hour
Wafers Processed
TXRF
Sample
PC & etch
50 Wafer
1
1
0
0
X
X
X
 
2
1
spiked
50
X
X
X
 
3
1
1
100
X
X
X
 
4
1
4
800
X
X
X
X
5
1
8
1500
X
X
X
X
6
1
24
4650
X
X
X
X
7
2
0
0
X
X
X
 
8
2
spiked
50
X
X
X
 
9
2
1
100
X
 
X
 
10
2
8
1550
X
X
X
X
11
2
8
4650
X
X
X
X
12
3
48
9300
X
X
X
 
13
3
48
9300
X
X
X
 
14
4
0
0
X
X
X
 

In the second experiment, the bath was spiked each day with 60 ppb of mixed ions. The same four analysis tools were used to determine the effectiveness of the purifier. The wafers were processed through the tool manually, so the etch times varied from run to run. The witness wafers used during this evaluation were immersed for about 10 minutes in the DHF solution. The tank's con centration was not changed at the conclusion of this experiment to determine if the purifier would precipitate metals back into the higher pH solution (see Table 1b for the testing schedule).

The purifier membrane was wetted with isopropyl alcohol (IPA) for over an hour and flushed with UPDI (ultrapure deionized) water, before being installed on the bench. After being installed on the bench, the process tank was flushed two more times with UPDI before introducing DHF chemical to the tank and purifier. The process tank was prepared by filling it with fresh chemical, performing a change out and then allowing the chemical to sit in the tank overnight. The next morning, the tank was again changed out to begin testing. A sample was pulled and witness wafers were processed through the freshly mixed chemical.

The bath was spiked with a known mixture of ions (Cu, Fe, Cr, Mg, Ni and Zn) in nitric solution. A 125 ml sample was pulled for ICP-MS (inductively coupled plasma mass spectrometry) analysis to determine the ion levels in the solution. ICP-MS analysis was used to determine trace quantities of metal elements in the chemical solution. The initial spiked sample was pulled from the bath, and the initial witness wafers were processed through the bath, while the pump was off after manually mixing the nitric solution with the DHF solution. The tank was static to prevent the initial spiked solution from being recirculated through the purifier.

Once the bath circulated the DHF chemical for one hour, another 125 ml sample was pulled for ICP-MS analysis and witness wafers were processed through the bath. Fifty 6750Å oxide wafers were processed through the DHF bath in 20 min intervals except during the processing of witness wafers and the collecting of chemical samples. The processing of these wafers to imitate product flow began after the sample was pulled and witness wafers were processed at hour one. Refer to Table 1 a and b for the process characterization schedule.

TXRF was used to determine metallic contamination from high molecular weight metals on the wafer surface that may have leached out of the chemical solution during processing. The TXRF monitors were bare silicon P-type prime wafers, and 2500Å thermal oxide wafers. The bare silicon wafers were placed in slots 1, 25, 50 during processing. The thermal oxide wafers were located in slots 13 and 38 in the reduce cassette.

The TXRF wafers were also used as particle monitors. All particle monitors received pre- and post-LPD (light particle defects) measurements that were carried out on a wafer surface analysis system, which has a 5 mm edge exclusion. The bare silicon wafers were measured at >=0.2 µm, and the thermal oxide wafers were measured at >=0.3 µm.

Thickness measurements were also conducted to determine if the purifier inhibits the etch performance. The oxide wafers were measured on a dual focus ellipsometer with a P/T ratio of <10% for thermal oxides in this regime. Nine points were measured on each wafer with a 6 mm edge exclusion.

DHF purifier setup

The Millipore Chempure disposable DHF purifier is a point-of-use purifier for dilute hydrofluoric acid (<1% by weight) that removes ionic contamination in a recirculating bath. The purifier is plumbed in series with the particle filters. Figure 1 shows the schematic of the recirculating bath system. The purifier pleated membrane cartridge has a composite membrane that contains a proprietary ion exchanger embedded in an ultra-high-molecular-weight polyethylene (UPE) matrix. Performance specifications of the cartridges include a maximum system pressure of 50 psi at 77 °F (25 °C) and a pressure drop/flow rate of 10 psi at 10 GPM. "The active sites, immobilized within the polymeric matrix, have a very high affinity for ions typically found in dilute HF solutions, including copper, iron, nickel, calcium, zinc, magnesium, manganese, cobalt and chromium. The metal ions are 'captured' by the active sites through a combination of ion exchange and chelation. Such a mechanism provides a stronger binding than ion exchange alone, especially in HF solutions where the complexing of metals ions with fluoride ions play an important role."5 Figure 2 illustrates the purifier's capture process of the ions, which is effective in solutions up to 1%HF.
Fig. 1. In a recirculating bath system, the dilute hydrofluoric (DHF) purifier is plumbed in series with the particle filters.
 
Fig. 2. In the first experiment, the bath was spiked each day with 30 ppb of mixed ions.

The purifier design has a high membrane packing density, high charge density, high mass transfer rates and fast kinetics at low fluid residence times. This high packing density does not affect the pressure or the flow.

Our evaluation was carried out with automated wet bench enclosed in a Class 1 mini-environment. The HF tank design has approximately a seven gallon capacity, while the recirculation loop and filters may hold up to four gallons of additional chemical. The recirculation module is composed of a pump, two filters, and plumbing lines that circulate the chemical and a heat exchanger. The heat exchanger is used to maintain the setpoint of the process chemistry.

The rinse tanks are configured as overflow rinse tanks (OFR). After the wafers were processed through the HF solution, the robot moves the wafers into the OFR, which is circulating the water through an inlet in the bottom of the tank. UPDI water flows into the tank at 14 GPM at ambient temperature.

The dryer system uses IPA liquid that is heated to a vapor state, then the vapor is forced into the chamber by flowing nitrogen (N2). The system is purged with N2 followed by a final dry.

Results

The ICP-MS data are the results from the samples pulled during test one and test two. The purifier removed more than 90% of the spiked ions in the first hour, except on day one of test 1. Refer to Figure 2 for test 1's results. Refer to Figure 3 for test 2's results. The graphs for test 2 were normalized to match the highest reading in test 1. Refer to Table 2 for an excerpt of the ICP-MS data from test 2. The data show the purifier removed 90%+ of the spiked elements from the bath within the first hour of purification and maintained the metal concentrations below 1 ppb for each element for several days. When fresh chemical is introduced to the membrane the ion levels are the same from concentrations in the previous bath. This is due to the low-level release of ions from the loaded purifier. When the purifier is removed from the system, the ion levels are lower than the system with the loaded purifier. This can be seen on Day 4 "no purifier" on Figure 3.
Table 2 Excerpt of the ICP-MS Data from Test 2
ICP-MS Results Test 2 (in pppb) Element
Day 2 Fresh HF
Day 2 Spiked DHF
Purified 1 hour
% Ions Removed
Day 3 Purified 50 hrs.
% Ions Removed
Day 4 Fresh HF Purifier
Day 5 Fresh HF No Purifier
Calcium (CA)
<0.10
12
0.72
94%
0.58
95%
0.52
0.1
Chromium (Cr)
<0.10
12
0.1
99%
<0.10
99%
<0.10
<0.10
Copper (Cu) <0.05
12
0.46
96%
0.54
96%
0.43
<0.05
Iron (Fe)
<0.10
12
0.12
99%
<0.10
99%
0.37
0.1
Nickel (Ni)
<0.05
12
0.91
92%
1.12
91%
0.85
<0.05
Zinc (Zn)
<0.10
12
1.21
90%
1.03
91%
0.78
0.11
 
Fig. 3. In the second experiment, the bath as spoked each day with 60 ppb of mixed ions.

TXRF data show the purifier reduces the amount of metallics in the solution, but unfortunately the purifier precipitates metals back into the solution until the chemical has been changed out. Even though the numbers are not catastrophic, they do raise some concerns including: 1) how much of the captured ions will precipitate back into the solution, and 2) how long is the purifier effective before beginning to precipitate these ions back into the solution?

The particle data below is an overall average of particle tests performed during both evaluations. The spec limit for this process is a delta of 50 adders at >= 0.2 µm. The particle data demonstrated no variance from run to run.

There is no significant variation in the etch process. The data below are from test 1. The wafers were immersed in DHF for 120 seconds, rinsed and dried. The oxide wafers measured were in slots 13 and 38. An analysis of average thickness and uniformity was performed for the DHF chemistry. The average thickness delta for the DHF chemistry was equal to 56.6Å. The within wafer uniformity was equal to 1.35%. The wafer to wafer uniformity was equal to 0.41%. The run to run uniformity was equal to 3.57%. The uniformity data were generated from the following formulas:

within wafer = (max-min)/(2*mean)

wafer to wafer = wafer max-wafer min/ 2*run average

run to run = (run max-run min)/(2*run mean)

These etch data points are comparable to the baseline etch performance presently seen on this tool.

Test 2 etch data were only collected to ensure HF chemical makeup during the evaluation. The data were within this facility's specification limits. Since the wafers were processed manually, the average thickness delta was not calculated for etch nor for uniformity.

Conclusion

The process characteristics used in this evaluation were ICP-MS, TXRF, particles and etch efficiency. These tests demonstrate that the purifier can reduce the ion levels in the solution by more than 90% in the first hour. Decreasing the ion levels in wet cleaning and etching processes by using the purifier can lead to significant improvements in wafer die performance, yield and lifespan. "Laboratory results have substantiated the significant adverse effects of chemical impurities on yield. The impact of unpredictable variation in ion levels will also increase as circuit linewidths decrease. Impurities not only corrupt thin thermal oxide layers, they increase wafers' surface roughness, which affects SiO2 dielectric characteristics. By keeping impurities to <1 ppb, significantly less surface roughening and better yield can result."6

However, the purifier does not solve the problem of having to continually monitor the bath for metal contamination. Since the purifier utilizes ion exchange technology, some of the captured metals may be released back into solution. The degree to which metal ions are released is a function of the solution pH and the amount of metal loading on the purifier. This can be seen on days four and five on Figure 2, which are HF sample at a concentration of 10:1. A lower pH solution of HF is used to clean the membrane once it becomes saturated. The disadvantage of using a purifier based on ion exchange technology is that without a constant monitor, the increased background level of ions due to purifier loading is unknown. The purifier does give process engineering greater peace of mind by minimizing the possibility of the process bath becoming contaminated. If a contamination spike does occur, the chances of affecting the product are greatly reduced by the purifier.

An ion exchange purifier is not ideal for "HF-last" applications. Since the wafer's been processed through SC1 and SC2, the metal build up in the HF over the lifetime of the bath is very low, typically less than 1 part per billion total. To be effective for HF last, a purifier needs to maintain metal concentrations at or near the levels in fresh Gigabit grade HF, typically less than 100 parts per trillion per element. This would require a chelating technology that provides "absolute capture" of metal ions.

However, an ion exchange purifier may be useful in HF first cleaning processes. Since wafers have not been processed through SC1 and SC2, the amount of metal added to the HF by each wafer is much higher, typically 3 to 5 parts per billion. Under these conditions, the background concentration of metals caused by the release of metal ions due to the reverse ion exchange mechanism is far less significant.   

Acknowledgments

Special recognition goes to Lisa Clark, John Rannells and Eleryn Ferrero for their assistance in making the ion solutions and in performing the ICP-MS analysis; Susan Backer for her assistance in the TXRF measurements; the APRDL pilot line team; and the managerial support of John Alvis, Roc Blumenthal and Fabio Pintchovski.

References

1. T. Hattori, "Trends in wafer cleaning technology," supplement to Solid State Technology, vol. 38 no. 5, p. S7, May 1995.
2. F. W. Kern, "Are ppt chemicals necessary?" Millipore 12th Annual Microelectronics Technical Symposium, San Francisco, July 18, 1994.
3. H. G. Parks, "Specification of liquid chemical requirements: The knowns, unknowns, and need for technology development," Semiconductor Characterization, Present Status and Future Needs, W. M. Bullis, D.G. Seiler and A.C. Diebold, eds., American Institute of Physics, 1996.
4. B. Parekh and J. Zahka, "Performance of a POU purifier in ionic contamination removal," Solid State Technology, August 1996.
5. B. Parekh and J. Zahka, "Point-of-use purification in DHF baths," Solid State Technology, July 1996.
6. P. Burggraaf, "New research shows, effects of chemical purity," Semiconductor International, p. 18, November 1993.

Darryl Weddington has been with Motorola for five years, the last three as a cleans process technician in the Advanced Products Research and Development Laboratory (APRDL). He is involved in developing, seeking, and maintaining wet clean process for 0.25 micron Logic technologies. He attended State Technical Institute at Memphis and St. Edwards University.

John Bodnar has worked in the Semiconductor Industry for 12 years as a wet process engineer, five at Texas instruments and seven at Motorola. He is currently an engineer at the Motorola's Advanced Products Research and Development Lab in Austin Texas.

Mo Jahanbani joined Motorola's Advanced Product Research Development Laboratory in July 1993 working on cleans development for advanced technologies. He is currently working in Dresden, Germany as a part of the Motorola/Siemens 300 mm joint venture project focusing on clean tools.


About SI  | Editorial Opportunities  | Editorial Calendar  | Media Kit  | Terms & Conditions  
Copyright © 2000 Cahners Business Information, A Division of Reed Elsevier, Inc.