Physical properties of specific chemicals create the basis for engineering calculations. Examples of simple physical properties include normal boiling point and molecular weight. Other commonly used physical properties may be more complex, such as specific heat and surface tension. Finally, more complex properties used for less common calculations include viscosity index, dielectric strength and others. One assumption is that these values are well known and accepted, at least for the most common and simple properties.

A recent experience tested this assumption. Recently, I started the work of converting some training material from Microsoft PowerPoint and Word for use in the open-source software LibreOffice Impress and Writer. After much evaluation the most secure conversion method was to manually re-create the documents. This way, no strange conversion artifacts end in the final work.

One part of this included a redoing of a handout table of physical properties including name, formula, molecular weight and boiling point. As this effort went along, a few of the numbers didn’t look as expected. A little digging into standard references revealed that, indeed, some numbers were inconsistent. The first thought was that cleaning this up should be straightforward. It wasn’t. Nearly four full days later I finished this “simple table.”

The inconsistency might seem minor, but it reveals a serious question: “Do we really know, to acceptable accuracy, the properties of the substances we use?” We use what we think we know for design, operation and troubleshooting. If we haven’t checked this, have we really done our job?

The starting point was looking up information in three major references: (1) the API Technical Data Book (4th edition), (2) the CRC Handbook of Chemistry and Physics (62nd edition) and Perry’s Chemical Engineers’ Handbook (6th edition).

The names got synonyms added as needed. In the U.S., H₂C=CH₂ gets called by the common name ethylene. In other regions, the more formal name ethene gets used. This really wasn’t a problem and made the final table nice and complete.

Fortunately, I didn’t need to update anything on the formulas.

The molecular weights came next. Unfortunately, the references couldn’t agree on this. Once you know the formula, getting the molecular weight is just adding up the atomic weights of the atoms. Finding values in the common references that didn’t add up to the values was a surprise. The errors even exceeded what you would expect from sloppy application of rules for rounding-off and significant figures. The differences were small, but having any differences at all was surprising.

Getting the normal boiling points straightened out consumed most of the time. Out of the 73 chemicals in the list, 44 had differences in boiling points. More than half of this list of chemicals had readily apparent inconsistencies. The chemicals here were not exotic materials rarely used in industry. Differences were found in boiling points for common and industrially important chemicals including methanol, ethanol and DMF (dimethyl formamide). Investigation showed that 12 of the 44 looked like they probably came from combinations of round-off and conversion between degrees Fahrenheit and degrees Centigrade. These errors were usually small. That leaves 32 more complicated situations.

How big were the differences? Let’s take two examples. First, for ethyl methyl ether, boiling points varied from 45.2 to 51.4°F (7.4 to 10.8°C), a range of 6.2°F (3.4°C). Second, for phosphine (PH₃) boiling points varied from -125.9 to -121°F (-87.7 to -85°C), a range of 4.9°F (2.7°C). This is at the high end of the scale. They are large enough that calculations using boiling points or properties derived from boiling points can be suspect.

The API Technical Data Book includes documentation for all its physical properties. If it included a compound, its properties could be traced and a decision made on the reliability of the value. In general, the default was that if the physical properties were in the API Technical Data Book, that was the value to use. Unfortunately, the API Technical Data Book was the least common of the references consulted. Many people lack access to it. Also, it’s not as complete as the CRC Handbook. Twelve of the 73 chemicals on the list were not in it.

The CRC Handbook was the most complete: 70 of the 73 chemicals were in it. For inorganic chemicals the tables in the CRC Handbook don’t reference back to the source material. For organics, the CRC Handbook cross-references back to sources cited in “Beilsteins Handbuch de Organischen Chemie” commonly shortened to Beilstein. You need to have a copy of Beilstein to trace back the physical properties to their original source. Not even all university libraries with chemistry or chemical engineering programs have copies of Beilstein or electronic access to a copy. There is some good news-bad news with Beilstein. All the first, second and third edition and some of the fourth edition volumes have fallen out of copyright, so you can find them on the Internet Archive. The bad news is that they are old enough that they have fallen out of copyright. Since Beilstein (like the others) is a compilation, the source information is even older. While updates have been added, the first volume of Bielstein’s fourth edition was published in 1909.

The consolidated CRC Handbook tables also have internal inconsistencies. Looking up “hydrogen, phosphide” listed as H₃P gives a normal boiling point of -125.3°F (-87.4°C). In contrast, looking up “phosphorus, hydride, tri” listed as PH3 gives a normal boiling point of -125.9°F (-87.7°C). Of course, H₃P and PH₃ are the same chemical. The different boiling points are in the same table. Even with the same reference, depending on how you look up the chemical, you may get a different value. It’s unclear if this is a typographical error, only partial updating when the value changed or evidence of using two different sources for the value.

Perry’s Handbook was a disappointment. In none of the cases where the boiling points in Perry’s differed from the others did the value in Perry’s physical properties summary (Table 3-1 and Table 3-2). hold up. Perry’s does include a second pair of tables (Table 3-7 and Table 3-8) of temperature versus vapor pressure. The temperature at 760 mmHg pressure is the same as the normal boiling point. Tables 3-7 and 3-8 appear to be more accurate. They also have the reference back to the 1947 summary tables they are based on. While you can use Perry’s, most people would default to the first set of tables (Table 3-1 and Table 3-2) and not realize the second set (Tables 3-7 and Table 3-8) give more accurate values.

How did the differences in boiling points get resolved? Nearly four days and another six references sorted out the confusion. Every boiling point in the table was referenced to the supporting source. The cross-referencing also noted if the original data was in degrees Fahrenheit or degrees Centigrade. Conversion round-off was correctly accounted for.

This effort may seem excessive for several reasons. First, this was a simple handout of limited use. After the training class, the attendees probably throw it away or file and forget it. Second, even simple calculations are often done with software with built-in values for chemical properties. So why even bother with this effort or write about it here?

The response is that you should never lose sight of the limits of our knowledge and what that implies. Do you know the limits of your laboratory reproducibility and repeatability? What physical properties, thermodynamic methods and other assumptions are built into the software you use? What is the accuracy and precision of the correlations being used? Being able to answer these is one key to effective troubleshooting. Understanding the degree of uncertainty you have helps evaluate if you should trust an answer or not.