Detection and Prevention of Ultra-Small Particles in IPA Vapor Drying
Technical Report 1103-21

Abstract:
Reduction in device geometries has resulted in product yield sensitivity to ultra-small particulate and contamination inherent in DI water, acids, and chemicals used in wafer cleaning. This contamination exists in the clean bath both as particles in solid form and as soluble substances that will precipitate out on evaporation of the water during the drying cycle. Since the resulting particles on the wafer are too small for detection on standard laser scanners (0.05 to 0.20 µ m diameter), a useful technique of coating the particles with a conformal CVD layer to accentuate the particles and make them detectable is reported. This technique has revealed that recent advancements in EPI’s IPA Vapor Dryer design has significantly reduced the number of ultra-small particles remaining on the wafer after the drying cycle.

Introduction and Objectives:
Progressive device geometries in VLSI and ULSI circuits have necessitated rigorous wafer cleaning technologies to remove chemical and particulate contamination prior to critical process steps. Over the last five years, device geometries have been reduced to the point where ultra-small particulate (0.05 to 0.20 µ m diameter) and contamination inherently found in the DI water, acids, and chemicals used in these cleaning steps can result in significant product yield loss.

Conventional drying technologies that spin-dry the wafers have been found to leave chemical and particulate contamination on the wafers due to a combination of residual left by evaporation of water and particulate introduced by electrostatic attraction. Even IPA Vapor Drying, if done without rapid recovery of the vapor zone¹ , can result in water mark formation that adversely effects product yield. These water marks consist of residual materials from the evaporation of water and localized oxidation.

Detection of ultra-small particulate on the wafer surface has been difficult due to the lack of metrology tools available to the user. Costly and time consuming tools that could determine the presence of small particles (e.g. SEM’s) offer no means of statistically trending performance or performing comparative analysis on process equipment due to the limited surface area that can be practically inspected.

This report details a processing technique that accentuates, or decorates, small particles on bare wafers and makes them optically visible to standard laser scanners used throughout wafer processing facilities. This affords the user the ability to statistically monitor the cleaning and drying process for these ultra-small particles and to perform comparative statistics on cleaning and crying equipment.

The objectives of this report are to describe the analytical technique used to determine the presence of small particles, and to report on a comparative analysis of EPI’s Yield Enhancement, High-Throughput, and Improved Design modifications (patents pending) to conventional dryer designs.

Experimental Method:

Experimental Design:

A simple three- factor factorial experiment was conducted as described in Table 1. 24 bare silicon wafers were processed for each CVD-Coating and Dryer-Design split in three separate runs over three consecutive days. The test wafers and all dummy wafers received standard RCA cleans immediately prior to each run. Full loads were used consisting of two 150 mm cassettes. The eight test wafers used for each run were placed in controlled positions throughout the load; remaining available slots were filled with the dummy wafers. Wet bench chemicals were changed at the beginning of each run. New IPA was added to each dryer at the beginning of the study and not replaced until the end in accordance with manufacturers specifications. Each dryer was cycled with dummy wafers processed through the RCA clean at least once per hour between test runs.

Dryer Design:
Two different dryer designs were used in this study. See Figure 1. The modified EPI dryer incorporated the Yield Enhancement, High-Throughput, and Improved Design Modifications (patents pending). These modifications serve to increase the bulk rate of heat transfer to the IPA by increasing the available heated surface area and increasing heater power. This allows a more rapid recovery of the vapor zone after insertion of the thermal load into the system. In addition, these modifications allow a much more evenly dispersed transfer of heat resulting in a more uniform vapor zone and the elimination of localized areas of rapid boiling or phase change.

Figure 1

CVD Coating:
After the wafers of each run were processed through the Vapor Dryer, half of them were immediately processed through a 900 Å LPCVD Phosphorous-doped Polysilicon process. Virgin control wafers were included with the test wafers from the vapor dryer splits to distinguish particulate contribution of the LPCVD and oxidation steps. The average particle level detected on the control wafers after oxidation was subtracted from the level measured on the vapor dryer test wafers (particles ranged from zero to eight particles per wafer on the control wafers).

The CVD Phosphorous-doped Polysilicon was then thermally oxidized for 15 minutes at 950°C. The oxidation served to prevent reflections from grain boundaries from interfering with detection of the small particles on the laser scanner.

Results and Discussion:
The results of the experimentation are summarized in Figure 2. The use of the CVD coating simply accentuated ultra-small particulate that was present on the wafers ran through the standard vapor dryer design. Although the use of the CVD layer revealed the presence of some ultra-small particulate on the wafer ran through the modified vapor dryer, the magnitude was reduced approximately 80%.

Measured Particles on Laser Scanner

Verification of the technique was accomplished by CVD coating, and remeasuring, the wafers from the first group (No CVD, Std. Dryer). After coating, these wafers gave identical results as the second group (CVD Coat, Std. Dryer).

Effects of CVD Coating:
The effect of using a CVD layer to high-light ultra-small particles can be seen in Figure 2. Comparing group 1 to group 2, it can be seen that an order-of-magnitude increase in the number of identified particles was seen using this technique on wafers run through the conventional dryer design.

By depositing a layer over the adhered particle the optical radius, and thus area, of the particle is increased allowing it to be optically detected by a laser scanner. Figure 3 shows the relationship between the optical radius of a spherical particle after deposition to the original particle radius and thickness of the conformal CVD layer.

Figure 3

Optical Radius (O.R.) = 23 (tr)

Optical Area = p (O.R.) ^ 2

For conformal CVD films deposited over spherical particles the optical radius is determined by:

R = r + t

O.R. = Ö (R² - (t - r)²)

= Ö ((r + t)² - (t - r)²)

= 2Ö (tr) for t ³ r

O.R. = r + t for t £ r

O.A. = p (O.R.)²

where: r: particle radius

t: CVD layer thickness

R: radius of the arc

O.R.: optical radius

O.A.: optical area

This technique is not limited to the Polysilicon CVD layers used in this study. Many different types of CVD films can be used for this application; often a suitable CVD layer exists as a standard flow in the wafer processing area. Use of Polysilicon usually requires a subsequent oxidation to prevent the grain boundaries from being erroneously detected as particles by the laser scanner. LPCVD nitride and oxides have been successfully used in this application and do not require reoxidation. PECVD films have also proven useful in this application as long as surface roughness remains low. It is essential that the CVD layer be nearly conformal or preferentially deposit on the particle.

Care must be taken when using the equations given above if the user is trying to determine the original particle size with this technique. Many CVD depositions will preferentially deposit on particles since they act as nucleation sites (particularly PECVD films). The conformal nature of all films can be different than expected when depositing over ultra-small particles of varying sizes and shapes. This does not distract from the ability to use this technique to determine the magnitude of particles on the wafer, but makes determination of the original particle size difficult.

Effects of Dryer Design:
As evident in Figure 2, the use of the modified Vapor Dryer significantly reduced ultra-small particles detected via use of the CVD coating. Two important observations can be made: A) the use of the modified dryer reduced particles by over 80% relative to the conventional design, and, B) the run-to-run upward trend of particles levels was reduced or eliminated with the use of the modified dryer.

The first effect can be attributed to the increased vapor zone recovery rate in the modified design. Quicker recovery of the vapor zone allows the particulate contaminate to be carried away with the water in solution before chemical bonding can occur with the wafer surface. Faster thermal recovery of the system also allows removal of the water before evaporation and corresponding precipitation of the contaminants.

The second effect is attributed to redesigns in the modified dryer that minimize ultra-small particulate buildup in the IPA and effectively block the particulate from being introduced to the vapor zone.

The rate of IPA conversion from liquid to vapor is determined by the rate of heat transfer to the IPA. However, local hot spots from poor heater design can result in vigorous boiling of the IPA. These hot spots also result in localized regions of superheated liquid and vapor IPA causing accelerated particulate buildup of polymers and long-chain carbon species. These particulate are entrained in the rapid mass transport near the hot spot of the IPA into the vapor zone.

The redesigned bottom heater of the modified EPI dryer provides a much more uniform transfer of heat devoid of localized hot spots. This drastically reduces the amount of particulate contamination in the sump. The condensate tray was redesigned to eliminate direct paths of vapor IPA from the sump to the vapor zone. The rerouted IPA loses sufficient velocity to entrain ultra-small particles. The particles fall back to the sump while the vapor IPA uniformly rises to the vapor zone.

Conclusions:
A technique utilizing CVD depositions has been determined to accentuate ultra-small particulate on bare wafer surfaces after clean and dry steps to make them detectable by standard laser scanners used in wafer processing facilities. This technique allows comparative statistics to be performed on cleaning and drying equipment.

Using this technique, a EPI Vapor Dryer with the High-Throughput, Yield Enhancement, and Improved Design modifications (patents pending) was compared to a conventional vapor dryer design. Results clearly show a significant reduction in particulate can be obtained with the use of the EPI system. Particulate can be obtained with the use of the EPI system. Particulate was reduced by over 80% with elimination of increasing particulate levels between IPA changes.

Post Script:
Information and data contained in this report has been provided to EPI by it’s customers under various nondisclosure agreements. As always, EPI upholds the utmost respect for customer confidentiality and has received permission from the specific customer to reveal this information in the format shown. In some cases the data has been normalized or similarly altered to prevent disclosure of proprietary information. This report was compiled and edited by the EPI Engineering Group. Any questions regarding this report should be directed to Kevin Schumacher of Essex Products International, Inc. At the time this article was published Kevin Schumacher was the Vice President of Operation and Engineering at S&K Products International Inc. He is now President and owner of Essex Products International.

References:
1. EPI Technical Report 1102-21: "Generation and Prevention of Water Marks in IPA Vapor Drying"