Such systems, despite their expense, were reasonably popular 15–20 years ago. However, for a variety of good reasons, which we will detail in this article, they quite justly fell out of favor, and essentially disappeared more than ten years ago.

To better understand this matter, it will be helpful to examine the regulatory history of ethylene oxide (EtO)…

In 1968, the American Conference of Governmental Industrial Hygienists (ACGIH) recommended a long-term exposure limit of 50 parts-per-million (ppm) as a time-weighted average (TWA), calculated over eight hours. In 1971, pursuant to 29 U.S.C. Sec. 655(a) (1982), which permits the promulgation of “national consensus standards,” the Secretary of Labor adopted the 50 ppm PEL (Permissible Exposure Limit).

By the mid to late 1970s, area monitoring systems for EtO were in place, based mostly on inexpensive, interference-prone detectors, which were hard-pressed to achieve even the modest sensitivity required.

In 1981, ACGIH lowered its recommended TWA to 10 ppm. Moving quickly, the organization designated EtO as a suspected carcinogen and proposed an even lower value of 5 ppm. In June of 1982, ACGIH recommended a TWA of 1 ppm, to take effect in 1984.

In January of 1982, OSHA issued an advance notice of proposed rule-making, inviting interested individuals to submit data or comments on revising the OSHA EtO standard. A year later, this led to OSHA publishing a proposed rule, suggesting a PEL of 1 ppm. After public hearings and some delay, in June of 1984, this PEL became a final rule.

Here are the current OSHA regulations, per 29 CFR:

§1910.1047(c)(1)

“8-hour time-weighted average (TWA).” The employer shall ensure that no employee is exposed to an airborne concentration of EtO in excess of one (1) part EtO per million parts of air (1 ppm) as an (8)-hour time-weighted average (8-hour TWA).

§1910.1047(c)(2)

“Excursion limit.” The employer shall ensure that no employee is exposed to an airborne concentration of EtO in excess of 5 parts of EtO per million parts of air (5 ppm) as averaged over a sampling period of fifteen (15) minutes.

Now, consider the implications. Although not explicitly stated in official OSHA guidance documents, the only way to determine an eight-hour time-weighted average of exposure is by first collecting full-time continuous monitoring data. The subsequent calculation can be performed manually, or as is more common, via specialized software.

Here is the problem with a gas chromatography/photo-ionization (GC/PID) system: Assuming that more than one sampling point is involved—and that is an excellent assumption in the SPD application—GC/PID provides but a single sensor, which must be time-shared among the various points. Thus, by definition, this is not continuous monitoring.

For example, on an eight-point system, a valve/timer apparatus conveys sample to the single detector, perhaps spending two minutes or so on each given point. In other words, in a 16-minute cycle around all of the points, only two minutes out of sixteen are being devoted to a given point. Therefore, no active monitoring data is being obtained for each point 87.5% of the time!

Yet, the purveyors of the GC/PID approach seem to embrace this deficiency, and attempt to transform it into a benefit. Superficially, you the customer are now able to “monitor” many points with one system. Ironically, the high cost (“sticker shock”) of a GC/PID monitoring system is lessened by adding more points—even though with more and more points, one is obtaining less and less data per point.

But, that’s not all.

Beyond the large safety issue raised by the monitoring system being offline for a given sampling point a substantial period of time, what about the accuracy and relevance of exposure calculations, based on such part-time monitoring data? In the eight-point system described above, how can the eight-hour average exposure be calculated?

Consider a given sampling point. As we have already established, data is recorded for only 12.5% of the time. Therefore, we must employ some sort of “placeholder” for the calculation. What should this placeholder be? How do we extrapolate the data?

One way would be to invoke the average value during those two minutes for the rest of 16-minute cycle. Another approach might be to invoke the highest value obtained. Yet another might involve some sort of trending algorithm, which seeks to anticipate future data.

Which method is correct? We at Interscan have no idea, and neither does anyone else.

While we’re at it, given these limitations, how in the world can an Excursion Limit (averaged over 15 minutes) be calculated? Simply put, under these circumstances, it cannot be done.

So much for GC/PID problems. You may well ask how these systems became popular in the first place. As noted above, early vintage EtO monitors were prone to cross-interference from many compounds that could be present in SPD departments. Despite the drastic change in regulatory standards, a goodly number of these original monitors were pressed into service for the 1 ppm compliance level.

Span adjustments were turned all the way up, and false alarms became the rule, rather than the exception. At about this time (late 1980s), PIDs were introduced that offered appropriate sensitivity, and when combined with a GC, offered specificity, as well. These systems were expensive, but the high cost could be amortized over the number of sampling points, even if this number had to be grossly inflated—based on what was actually required.

Thus, the GC/PID systems were deployed. Most of the time, they appeared to be doing a fine job, and were standing up to all the interferences, including isopropyl alcohol—the most notorious.

On the other hand, GC column replacement was costly, and if there were enough isopropyl alcohol, it could “swamp” the column, so the interference and false alarm would occur anyway. Recall that PIDs are not at all specific in themselves, and must rely on a GC column to achieve specificity.

In addition, end-users were noticing that alarm events were sometimes being missed during the off-time of a particular sampling point. A few end-users even realized that the data acquisition reporting left much to be desired.

By 2004, or thereabouts, it would be difficult to find a GC/PID system in the SPD application.

We can explain the mini-resurgence of GC/PID by noting that biases seem to exist in all fields of endeavor, and the world of analytical instrumentation is no exception. In the laboratory, GC is considered an “elegant” method. In this case, though, such elegance disintegrates under the demands of workplace occupational health monitoring.

It would be most unfortunate if this notion of analytical elegance blinds specifiers to the quite substantial limitations such elegance might incur. These limitations must not be imposed on vulnerable workers in SPD departments.

Everyone around the world agrees that a part-per-million (ppm) equals 10^–6. Unfortunately, since everyone around the world does not agree with what a “billion” is, they can’t agree on what a part-per-billion (ppb) is, either.

In what is termed the American system, 10⁹ is a billion. But, in the European system, 10⁹ is a milliard (sometimes called thousand million). In the Euro system, a billion is 10¹². Ironically, the French used what is now the American system as far back as the seventeenth century, before switching to the Euro system in 1948. Are you confused yet?

While I have never encountered the term parts-per-milliard to signify 10^–9, it certainly could happen. Imagine if that were also abbreviated “ppm”!

However, the unit µl/M³ [microliter per cubic meter] has been published in numerous official Korean documents, and was used to avoid the potential parts-per-billion confusion. This interesting unit does define 10^–9 in a non-ambiguous fashion.

The good news is that American cultural hegemony has already affected the numbering system, and, as you might expect, money was involved.

We prevailed upon (at least) the Brits that 10⁹ dollars be called a billion dollars, and this was agreed to in 1974, under Prime Minister Harold Wilson. This usage has also spread into common parlance, and according to the The Times of London style guide,

[a] “billion [is] one thousand million, not a million million”

Even though our definition of a billion as 10⁹ is supposed to apply in official government statistics worldwide, the Korean example is clearly an exception, and is probably not unique.

Our recommendation is to ask what someone means by parts-per-billion, in any international dealings.

As our hero was slogging through some e-mail inquiries, he came upon two units of measurement he had not seen before:

ppmv   and    µg/Nm³

ppmv

ppmv is simply parts-per-million by volume. This notation would distinguish it from parts-per-million based on weight. As it happens, when one is working with solids or liquids—i.e. when one leaves the world of gas measurements—parts-per-million is understood to be by weight (strictly speaking by mass).

As discussed in the article referenced above, the most proper way to think of parts-per-million in the gas world is as molar concentration: µmoles of minor component / total moles in mixture. By tradition, and because ppmv would be the same as the molar concentration for ideal gases, many people still speak of gas measurements as being in parts-per-million by volume.

µg/Nm³

µg/Nm³ means micrograms per normal cubic meter (Nm³). The “normal” cubic meter is defined as being at 0°C (273.15°K) and 101.325 kPa or 760 mmHg (i.e. 1 atmosphere of absolute pressure). However, this notation is no longer appropriate unless the specific reference conditions are explicitly stated, since there are currently many different definitions of what constitutes standard reference conditions.

Here are the full names of the entities listed in the above table–

As you can see, by using the unit µg/Nm³, you are bound to be misunderstood—if not in the definition of normal or standard conditions, then by the difficulties inherent in using mass/volume units rather than parts-per-million.

This danger may result from some sort of upset condition, as in a leak—representing an immediate toxic or explosive risk—or it could be more insidious, as in a long-term relatively low-level exposure to some toxic compound, devastating only in its cumulative effects.

Although many compounds list both occupational exposure limits and lower/upper explosive limits, in practice, there are few common hazardous gases that will present themselves as both toxic and combustible dangers. One of these is ethylene oxide (EtO). Even though the widest use of this chemical is in chemical syntheses, in terms of numbers of people affected, the health care related applications are of the greatest interest.

A recent case serves to illustrate the dangers of EtO, and what can happen if those concerned—both on the user and regulatory sides—lose sight of this.

EtO is an essential sterilant, used on the many medical devices that cannot take the heat of steam. Despite the introduction of numerous processes that promised to replace EtO, none has. Indeed, Johnson & Johnson is still one of the world’s biggest health care consumers of EtO even though it also markets a sterilizer touted as a partial EtO replacement.

On 19 August 2004, at a contract sterilization facility in Ontario, California, mishandled EtO caused an explosion, that resulted in injuries to four employees, and damage to the facility that disrupted normal operations for nine months.

Brilliant and detailed investigative work on this incident was performed by the U.S. Chemical Safety and Hazard Investigation Board (CSB).

CSB’s webpage on the investigation is here

A very informative video, featuring forensic-quality animation, is available for live-streaming.

DVD copies of this program, and others created by CSB, are available on request.

After studying CSB’s materials, one will note that:

1. In general, all concerned seemed to be detached from the reality the EtO is explosive. During a maintenance procedure, a key aeration phase was bypassed, and shatter-resistant glass was not utilized in the control room windows overlooking the process. Indeed, every injury was caused by flying glass.
2. System designers were under the quite mistaken belief that monitoring pressure can substitute for gas detection.
3. In its zeal to protect us from “evil” EtO, California and certain other states continue to require the treatment of backvent EtO emissions, even though the Federal EPA (hardly an anti-Green organization) eliminated this requirement in 2001. That catalytic oxidizers—featuring live flames—are needed in this process doesn’t seem to bother anyone.

Therefore, in the minds of regulators, the very questionable “hazard” of venting EtO to the atmosphere, as could have been treated by the much safer chemical scrubber method, is somehow more important than the obvious risks of employing the catalytic method, which, in fact, catalyzed this explosion.

Lest we forget, the biggest danger involved in this entire enterprise is that if not properly sterilized, the medical devices can cause serious illness or death to hapless patients! Indeed, if EtO were used more, and the various alternative methods were used less, we could probably put a dent in the more than 100,000 deaths caused by in-hospital infections, that occur every year in the United States.

It is also well worth remembering that with technological solutions do come some potential dangers. No one argues that we should return to the horse and buggy, even if the gasoline in automobiles is dangerous. Users of gasoline are cautious, and understand the hazards involved.

Similarly, sterilization is a boon that drives beneficial invasive medical care. EtO and other potentially dangerous chemicals can be handled safely, but only if both the users and the regulators understand the dangers, and impose logic, right reason, context, and perspective on the methods involved.

The “no safe level” theory most likely got its start in the nuclear industry, where it is called the Linear-Nonthreshold Dose-Response Model (LNT). This model holds that any exposure to ionizing radiation (other than zero) is harmful, in that damage to a single cell could be sufficient to cause cancer. And, the higher the dose, the more extensive the damage. As such, there is no “threshold” level below which damage would not occur.

LNT was evaluated by the National Council on Radiation Protection and Measurements in 2001, and they determined that:

“…there is no conclusive evidence on which to reject the assumption of a linear-nonthreshold dose-response relationship for many of the risks attributable to low-level ionizing radiation although additional data are needed. However, while many, but not all, scientific data support this assumption, the probability of effects at low doses such as are received from natural background is so small that it may never be possible to prove or disprove the validity of the linear-nonthreshold assumption.”

Note here that the natural background is acknowledged, and that proving (or disproving) the theory is probably not possible. By acknowledging that a natural background level of radiation exists, the necessary conclusion is made that our bodies can and do deal with these small levels, such that no harm is actually caused. Alternatively, one could posit that even background levels of radiation—which we are all exposed to—can cause cancer. I would reply that if that is the case, why worry about any other sources?

A riff on this is toxicologist Robert Golden’s argument against LNT:  If there is no safe level, why bother doing any health effects research at all?

Lead is regulated under CPSIA as follows:

• Beginning 180 days after the date of enactment of this Act (10 February 2009), the lead limit is 600 parts per million total lead content by weight for any part of the product.
• One year after the date of enactment of this Act (14 August 2009), the lead limit is 300 parts per million.
• Three years after enactment of this Act (14 August 2011) the lead limit is 100 ppm (if technologically feasible).

Lead can be extremely damaging to developing brains—especially in utero and many shrill commentators, speaking in support of CPSIA, note that “There is no safe level of lead.” Yet, the very law they champion does, in fact, specify a succession of safe levels, and even allows the possibility of waiving the 100 ppm requirement, should it not be feasible. Instead, the Consumer Product Safety Commission must then establish the lowest feasible amount below 300 ppm, and review this finding at least every five years.

But, this is only logical. No law can be written proclaiming a standard of zero. As measurement technology develops, the notion of “undetectable” gets smaller and smaller. Enforcement must always be based around a measurable standard.

More than that, “no safe level” cannot possibly be applied in any other real life situation. For example, under a regime of “no safe level,” no pharmaceutical drug could exist, as nearly every drug has some dangerous side effect that could affect one out of a million people.

People drive all the time, yet driving is far from 100% safe. Should parents stop allowing their kids to participate in sports given the very real possibility of injury?

The only appropriate use of “no safe level” would be in this sentence: There is no safe level of the public being exposed to any commentary that includes “no safe level.” Remember that “no safe level” inspired the draconian mess that is CPSIA.

As always, good science plus common sense equals good public policy.

For example, thermal flowmeters are based on convective heat transfer effects, and can be calibrated at ambient conditions on a specific gas, and then be used at process conditions to make accurate mass flow measurements. However, if the process gas is highly reactive or toxic, it may be difficult or impossible to perform a calibration, even at ambient conditions.

In such cases, it is common practice to calibrate the flowmeter with a substitute (surrogate) gas, that matches the thermal characteristics of the target gas as closely as possible, while acknowledging that error may be introduced.

In the world of LEL (lower explosive limit) gas detection, catalytic-bead pellistor type sensors respond to many combustible gases. Ideally, they should be calibrated with the target gas—if it can be identified. Sometimes, though, this is not possible, and methane, propane, or pentane standards are used as surrogates. Fortunately, relative response data, based on how the sensor would respond to a given target gas—calibrated with each surrogate—is available.

As it happens, calibrating with pentane causes enhanced response for most gases, methane causes a lowered response for most gases, while propane strikes a middle ground.

For toxic gases, calibration standards—either cylinder gas or permeation devices—are available in virtually all cases, and surrogate calibration is not needed. However, this has not stopped certain gas detection companies from advocating surrogate calibration for what one might call “convenience” reasons. They may advocate the use of a carbon monoxide standard to calibrate an ethylene oxide instrument since CO standards are cheaper and easier to come by, even if ethylene oxide standards ARE readily available.

There are some dangers here, though. At best, the surrogate conversion factor (1 ppm surrogate registers x ppm as the target gas) is a single number, based on testing of new sensors, maybe with a fudge factor built in to adjust for how this ratio changes with sensor age. Maybe. The problem is that no one can tell you at any particular time how YOUR sensor, installed in the field, and calibrated with a surrogate, will react to the target gas.

As you might suspect, such uncertainties do not bode well if litigation should arise. Be assured that plaintiff’s lawyers know all the right questions to ask about calibration and maintenance of gas detection systems.

Best practice calls for a rigorous chain of custody of calibration, traceable all the way back to some recognized standards body, such as NIST in the United States. It is unlikely that this can be achieved with surrogate calibration.

In those very rare instances where a surrogate calibration is necessary, a responsible manufacturer will disclose all the details and pitfalls.

Pioneering work by NCASI, the The National Council for Air and Stream Improvement, had pegged the response ratio such that approximately 3 ppm of chlorine is required to cause a 1 ppm reading on an Interscan chlorine dioxide analyzer.

Subsequent efforts have refined the number a bit. We now advise our customers that:

2.8 ppm of chlorine will show a 1 ppm reading on a chlorine dioxide analyzer, and this ratio is stated with an accuracy of ±10%

You are correct in noting that the response of our instruments is linear. More than that, electrochemical voltametric sensors (such as we use) are inherently linear, with no electronic compensation required.

However, for optimum accuracy in most applications, it is best to calibrate the instruments at a concentration somewhere around 50% of the scale range or higher, if you can. This is because there will always be various sources of error beyond the sensor electrochemistry, and prudent analytical technique would frown on calibrating an instrument at 1% of the scale range for readings at 90% of the scale.

Furthermore, for optimum accuracy, if for some reason you must operate toward the bottom of the range, ideally you should calibrate close to this level.

Saying that, for most customers, calibration at 50% of the range or higher should be quite satisfactory.

For more information, please consult our articles on accuracy and minimum detectability.

Rise time to 90% of final value, rise time to 50% of final value, and fall time to 10% of original value are given for all gases, and specialized sensor types for hydrazine and hydrogen sulfide. It is noted there that the 50% figure is useful when considering how fast an instrument will respond in a critical (alarm) situation, whereby “full” response is not as important as the “step” response to an immediately hazardous concentration of toxic gas.

It is further noted that sensor rise and fall times are affected by many factors, including age, chemisorption, cumulative exposure to target gas and interfering gases, and maintenance. Data is also provided to help the user determine the lag time caused by interconnect tubing used to draw sample in from remote points. Fortunately, even 100 feet (30.48 m) of typical ¼ inch (6.35 mm) OD tubing introduces a lag time of only 29 seconds.

It is important to temper the sometimes misplaced zeal for fast response times, and zero lag intervals.

When real-time gas detection instruments were first introduced, it was natural to compare their relatively instant response with the predominant wet chemical or detector tube methods. As more instruments came on the market, using different operating principles, one of the specifications that was inevitably compared was response time. “Lag time,” used in a context where there is no interconnect tubing to a remote point, refers to whatever inherent system delay exists, before the sample gets to the detector. Generally invoked for instrument methods that have quick detector response times, but plumbing issues that slow the overall response, this definition of lag time was judged to be “more fair” to such techniques, but becomes a pointless distinction to an instrument user.

Moreover, lightning fast response and fall times may look good on a brochure or website, but offer few real-world advantages, beyond allowing a particular type of detector to be used in a stream-switching context with more points than a slower responding detector might allow. Precious few situations exist whereby an instrument that responds a few seconds quicker can be said to offer any advantage. Nearly all sample-draw direct-reading toxic gas analyzers—such as Interscan’s—respond fast enough for 99% of applications. Saying that, if one is considering a diffusion sensor approach (no sample-draw pump) one definitely SHOULD take into account the inherent lag time of such devices. There are many factors to consider in this case, including ventilation characteristics, measuring range, and reactivity of the target gas.

As to the matter of lag time due to interconnect tubing, this is of practical significance only in stream-switching installations. There is a valid concern here that since data is not continuous for all points all the time, the newest and most updated sample should be quickly presented to the instrument. However, in a continuous monitoring installation, few situations are so critical that adding 30 or so seconds of lag time would matter. And, if it does matter, steps can be taken to minimize even this small amount of lag time.

Please remember that as important as instantaneous high concentration alarms may be, most occupational health monitoring is ultimately more concerned with long-term exposure, based on an 8-hour time-weighted average. Even the NIOSH IDLH (Immediately Dangerous to Life and Health) levels are based on a 30-minute exposure, and best practices for any monitoring system would take into account either the short term exposure limits of 15 minutes, or the ceiling value (if the target gas has one) that requires the fastest monitoring.

In other words, diligent applications engineering and proper design of the gas detection system in the first place will make any discussion of lag times caused by interconnect tubing to be moot.