Use of ID Pressurization to Control Weld Profile in Semiconductor Ultra-High-Purity Process Gas Lines

Use of ID Pressurization to Control Weld Profile in Semiconductor Ultra-High-Purity Process Gas Lines

Welding of high-purity process gas lines for the semiconductor industry is the largest single application of automatic orbital GTA welding. Several millions of welds have been done using this technology since it was introduced to the semiconductor industry in the early 1980's. Nevertheless, there are still differences in methods and techniques used by various companies and contractors to achieve the goal of an ultra-high-purity gas delivery system. The conventional weld for 1/4 inch diameter (OD) tubing is a single-pass weld with pulsed current. Dr. Tadahiro Ohmi and his colleagues at Tohoko University in Japan have done extensive experimentation developing a technique of multi-pass welds without the use of pulsation that received considerable attention several years ago. One of his techniques involved the use of a Magnehelic pressure gauge to determine when the fit-up between the tubes being welded was correct.

Recently, Mr. Jim White of Dynamic Systems, Inc. (DSI), a mechanical contractor that installs high-purity piping systems for semiconductor process gas lines, has experimented with the use of pressurization during welding to control the inner weld bead profile. Although the use of pressurization to control the weld bead is not particularly new, Jim has provided documentation of the specific amounts of pressure required as a function of tubing diameter and length. He has supported this with results from the installation of process gas lines at several Texas semiconductor plants during which thousands of welds were completed.

Typical tubing diameters for semiconductor process gas lines are 1/4, 3/8, and 1/2 inches, with 1/4 inch being by far the most common size. Although 1/4 inch is the most frequently used, it is also the most difficult size with which to achieve consistent repeatable high-quality welds. With a typical wall thickness of 0.035", one ampere of welding current can have a large effect. Since every heat of material has a slightly different chemical composition, the heat input, and thus the welding current, required to weld each different batch of tubing can vary by as much as several amperes. Striving to achieve a flat exterior bead with no concavity by reducing the amperage can result in a lack-of-fusion defect on the interior weld bead. This means that the weld schedules or parameters must be changed whenever tubing heat numbers are changed. Since it is common to have several different heats of material in use at the same time, this means that welding operators must adjust to the heats which are being welded and change the weld schedule accordingly. Weld parameters must also be changed when welding dissimilar heats or when welding tubes to fittings, valves, or other piping system components. DSI has found that by using a weld program in which the applied heat (amperage) would result in overpenetration and OD concavity in the absence of pressurization, that pressurization can be used to control penetration. With pressurization, the same weld program can be used for most material heats and fittings of a similar size and wall thickness so that the weld program does not have to be changed each time. The weld bead would be narrower on fittings and valves than on welds of tube-to-tube.

Fig. 1 Tee-fitting used to measure ID pressure at weld joint prior to welding. Plastic tubing leads to a Magnehelic gauge. Fitting is removed and replaced with weld components to make the weld.

To determine the correct amount of pressurization, DSI installs a SwageLok® tee in place of the weld head. (Fig. 1) One branch of the tee is connected to the Magnehelic gauge to determine what the pressure will be at the weld. The tee is then removed from the system and the weld is accomplished.

Pressure variations occur during the welding process. When two tubes are installed in the weld head in preparation for welding, there will be a slight gap between them sufficient for some purge gas to pass through. At the start of the weld there will be some shrinkage as the weld puddle is established causing the tubes to pull slightly apart. This causes a small pressure drop to occur. As the weld progresses, the tube ends are drawn back together until the weld closure after one revolution. The pressure measured at the tee prior to welding will be equal to the maximum pressure attained at the completion of the weld. Ohmi used this observation to get optimum alignment of components prior to welding. The lower pressure during the weld is sufficient to prevent concavity. The maximum pressure will not cause convexity or blowout. (Fig. 2, 3 & 4).

Pressure results from I.D. purge gas flow rate (CFH) and the size of the restriction orifice at the end of the tube (or tube extension). A 1/8" diameter exit orifice is used on tubing up to 3/4" OD and pressure is adjusted by increasing or decreasing the flow rate. DSI has found that valves have flow restrictions that are not necessarily uniform from piece-to-piece. In those cases it is necessary to measure the pressure for each part before welding. When welding fittings or tubes, which provide very consistent amounts of flow restriction, reductions in flow rate can be made based on previous measurements rather than requiring a pressure measurement for each weld. DSI has developed a sense of how much flow rate reduction is needed to weld a pipe 10 feet long compared to welding shorter pieces at the bench. Since the velocity of gas flow is much greater in small diameter tubing, a small change in the flow rate will have a much greater effect on the pressure in the smaller tube.

Fig. 2 Welding technician in the DSI cleanroom setting up a weld on the Model 9-500 weld head. Model 207 power supply is installed under the shelf while a chart recorder on the shelf monitors weld parameters.

Fig. 3 When welding in the DSI cleanroom, the use of gloves and cleanroom clothing is mandatory. Here a welding technician prepares to weld with the Model 9-500 weld head.

Fig. 4 Welding set-up in DSI cleanroom:
Model 207 power supply, Model 9-500
weld head, remote pendant at center, right.
Magnehelic gauge is mounted on the wall.

When pressurization is used to make a single-pass weld with pulsed current, the resulting weld bead is wider than when no pressure is applied, and the outer weld bead surface can easily be made flush with the tubing surface. A convex outer weld bead would indicate an excessive amount of ID pressure and a concave inner weld bead which is unacceptable. Too much pressure can result in the weld "blowing out". This is because the liquid weld puddle is held in place by surface tension. It takes only a slight amount of pressure to move the weld puddle outwards causing the molten puddle to come in contact with the tungsten electrode and make a hole in the weld.

The amount of pressure required to achieve an ideal weld bead profile varies with the tubing diameter. For 1/4 inch tubing, 3-1/2 inches of water column (W.C.) is typical. Less pressure is used for larger diameters, such that for 3/8 inch tube 2-1/4" W.C. and for 1/2 inch tubing, 1-7/8" W.C. is required. For tube sizes greater than 1/2 inch diameter, only a minimal amount of pressurization is applied (Table 1).

Weld Bead Width

Weld bead width is somewhat controversial. Sometimes welding specifications will indicate that the inner weld bead must be no wider than one "t" where "t" is the wall thickness of the tubing being welded. The reasons for this are partly cosmetic, but Ohmi showed that the smoother the ID surface of tubing or components in a high-purity gas piping system, the less likely the possibility of contamination. A rough surface is more likely to absorb moisture or to entrap particulates that could contaminate a process and reduce product yield, thus the wide-spread use of electropolished tubing. The weld, however, has a much rougher surface area than the rest of the system, so it would be logical to try to keep the weld area as small as possible. This has driven some end users to demand very narrow inner weld beads.

In our opinion, a weld program which uses minimal heat in an attempt to obtain a very thin weld bead is less likely to provide uniform consistent welds, and more likely to result in lack-of-fusion than a slightly hotter program which will be more forgiving. A very thin weld bead can be obtained with consistent results using weld programs with higher electrode travel speed and multiple passes as shown by Ohmi and his co-workers. A thin weld bead requires a consistently perfect fit-up in order to get repeatable good results, but a wider weld bead is more forgiving and somewhat less dependent upon fit-up. A wider weld bead greatly reduces the chance of a lack-of-fusion defect.

This type of defect provides a site for particulate entrapment and provides a corrosion initiation site. Any lack-of-fusion on the weld ID is cause for weld rejection. There is no common agreement on weld bead width with one semiconductor manufacturer showing a preference for wider weld beads of one to three times the wall thickness and another preferring narrower weld beads. Specifications may differ somewhat for visual weld criteria at each particular job and these specifications don't usually define a particular weld bead width as long as penetration is complete and the weld bead is uniform around the weld joint circumference.

The width of weld bead relative to the depth of penetration is called the W/D ratio. This ratio is much greater for heats of stainless steel with sulfur contents less than 0.008%. (See Fig. 5). Above this level of sulfur the weld bead shows a greater amount of penetration relative to the width of the weld. When welding material heats of different sulfur contents, with one heat below 0.008% and the other heat above this level, the arc often deflects towards the heat with the lower sulfur content. This can result in a weld bead that is entirely on the low sulfur member and a failure to get complete fusion of the inner weld bead.

At the Sematech facility, DSI used two passes to successfully weld a low sulfur tube to a higher sulfur valve. They developed a two pass weld procedure which was approved by quality control (QC) and certified for the particular heat numbers involved. The procedure had to be requalified if the heats were changed. In one instance, the mismatch between the two material heats was so bad that they had to allow the weld to cool completely after the first pass before completing the second pass in order to get a successful weld.

Process gas piping systems have been installed by DSI using the weld pressurization technique at Cypress Semiconductor in Round Rock, Texas, National Semiconductor in Arlington, Texas, and Hitachi Semiconductor in Irving, Texas, at VLSI and Sony in San Antonio as well as at the Motorola facilities in Austin, Texas and Phoenix, Arizona. In these facilities, all specifications for industry criticality for UHP process gas piping systems have been met. Trace gas impurities have been held to single digit parts per billion (ppb) levels while particle counts have been limited to single digit levels per cubic foot for particles of 0.014 up to 3 microns in size as measured by a CNC Condensation Nucleus Counter.

Table 1. ID Pressure (W.C.)
Applied as a Function of Tubing Diameter
Tubing DiameterID Pressure
1/4 inch3 1/2 inches
3/8 inch2 1/2 inches
1/2 inch1 7/8 inches
>1/2 inchminimal pressure

Pressurization Techniques

DSI indicates that their ID pressurization technique is much more practical for fabrication and field installations than the conventional technique employing multiple programs for a single tube diameter. Jim White has personally done 4,000 to 6,000 welds with the ID pressurization technique. He states that blowouts have been completely eliminated. He has 50% fewer rejects and better overall productivity when using the pressurization technique. Jim says that "when working in a fab shop, anything that can be done to make life easier is money in the bank. "

AMI's Findings

As a welding equipment manufacturer, AMI finds that our customers have developed and use several techniques or welding methods for achieving UHP gas piping systems. Single-pass welds with pulsation with or without ID pressurization and multi-pass welds with or without pulsed current, and with or without ID pressurization have been used successfully by different customers. Excellent consistent welds are achievable with each of these techniques.

The difference between a single-pass weld and a multi-pass weld is simply a matter of programming the power supply to turn the weld head rotor a sufficient amount of time to complete either one, two, or more passes at the desired RPM.

In general, a multi-pass weld would be done at a faster travel speed. In pulsed-current welds, the primary amps are higher than for non-pulsed welds because of the cooling effect provided by pulsation. Non-pulsed welds are not recommended for larger diameters and heavier wall thicknesses (Fig. 2 & 4).

In order to achieve consistently high standards with any welding technique, a well-defined set of weld criteria should be in place. It is also important that welding personnel are trained and experienced with the particular welding technique and finally, good quality control is essential.

On the issue of contamination in welding systems, in discussions with facilities managers and personnel, we have not been aware of any problems related to properly done welds with any orbital welding technique. We can cite examples of ultra-high-purity gas distribution systems that meet or exceed current standards of gas purity installed with each of the welding techniques mentioned above. If there is an advantage in this respect of one welding technique relative to another it is very minor compared to the contribution of particulates and other contaminants in such systems from valves and other piping system components.

By Barbara K. Henon, Ph.D., Arc Machines, Inc., and Jim White, Dynamic Systems, Inc.