Grantville Gazette, Volume 72

by Bjorn Hasseler

  Airship lift depends on the difference in density between the lift gas and the ambient air, and thus in part on their respective temperatures. So temperature is relevant to airship pilots, not just farmers.

  We need to start making accurate records of weather conditions, and we will certainly be looking at temperature, humidity, and precipitation.

  ****

  Temperature

  In Grantville, the most common form of outdoor thermometer is the liquid expansion thermometer. Most such thermometers will probably use, as the "thermometric liquid," an organic liquid with a red or blue dye, but it will be common knowledge that mercury may also be used.

  Mercury has the advantages of being opaque, easily purified, chemically stable, not wetting or chemically attacking glass, liquid over a wide temperature range (-38.8oC to 356.7oC, thus unlikely to evaporate at the top of the column), and having clearly defined meniscus, a high thermal conductivity, low specific heat (making it rapidly responsive to changes in temperature), and a fairly linear coefficient of thermal expansion. Unfortunately, it is poisonous, the expansion is small compared to alcohol, and in very cold climates it can solidify.

  Several different organic liquids have been used, but the most readily available in the 1632 universe is ethanol, with a liquid range of -114 to 78oC. My prediction is that mercury will be used for a small number of precision reference thermometers and the actual weather stations will use ethanol thermometers.

  Glass composition is also significant. It is not just the liquid, but also the glass, that expands as the temperature increases, and not entirely linearly (or at the same rate as the liquid). Also, after being first heated and then cooled, the glass bulb of some compositions did not return to its original dimensions, leading to a slow rise in the zero of mercury thermometers. In the late nineteenth century, Schott developed a series of more stable glasses, notably borosilicate (Pyrex(R) glass) (Vogel 21). Hard glasses are generally preferred (EB11/Thermometry).

  Most modern meteorological thermometers have the stem, with engraved scale markings, inside a protective glass sheath, and there is a white enamel backing on the stem to make the liquid movement more visible. (Srivastava 96). My "hardware store" thermometers are unsheathed, and the scale is on a separate attached metal frame. The attached scale will "inevitably move slightly with time." (Burt 116).

  The first thermometers sensed air rather than liquid expansion. The first known drawing of a thermometer is from 1611. It shows an inverted flask with a long narrow stem, fitting into the neck of a short-necked flask, the latter partially filled with water. The bottom of the stem of the first flask is below the liquid surface. A rise in temperature caused the expansion of the air in the short flask, pushing the water up the stem. Alongside the stem there was a scale divided first into eight degrees and these each into six ten-minute intervals. Its inventor, possibly living in Rome, is unknown (Middleton 11).

  The basic problem with unsealed air thermometers was that the expansion of the air was a function of pressure as well as temperature. In 1632 Jean Rey (1583-OTL c1645) dispensed with the second flask, and turned the first flask stem upward, creating a liquid expansion thermometer. However, the tube was unsealed so errors could arise from evaporation of the water (27). The sealed spirit-in-glass thermometer is attributed to Ferdinand II, Grand Duke of Tuscany, and most likely invented in 1654. The first experiments with mercury were in 1657, but the Tuscan academicians deemed it inferior in performance (28-37).

  Before leaving the subject of early temperature measurement, I wish to call the reader's attention to Fitzroy's chemical weather glass (1862), as is it the sort of curiosity that a resident of Grantville might have inherited, or picked up at a craft fair, before the Ring of Fire. It "consisted of a solution of camphor and certain inorganic salts in aqueous alcohol, sealed in a glass tube." Negretti & Zambra used potassium nitrate and ammonium chloride. The salts formed crystalline dendrites, and Fitzroy claimed that when the crystals built up, the weather would get colder and stormier, and if they disappeared, it would be dry and clear. Studies by Mills have shown that the chemical weather glass is sensitive both to the current temperature and "any preceding regime of temperature changes." It is thus a thermoscope. Mills comments, "A rapid fall in temperature associated with an approaching vigorous cold front could conceivably trigger ... rapid crystal growth if observed at a fortuitous time, but in general any correlation between appearance and future weather patterns would be purely coincidental."

  Manufacture. In 1612, Giovanfrancesco Sangredo (d. 1620) made several thermoscopes, at a cost of four lire each. These had no scale, but the column height could be measured with a caliper. He apparently made use of "a wine glass with a foot, a small ampoule, and a glass tube," and he could make ten in an hour.

  The Grand Duke's glassblower, Mariani, had incredible skill and was able to manufacture thermometers with a "50 degree range" (corresponding to the modern -18.75 to 55oC) with great consistency. He admitted, however, that he could not do this for the Medicean 100- and 300-degree range thermometers, because "inequalities could more easily occur in the larger bulb and longer tube" (Middleton 34-5). On the other hand, Middleton asserts that "workmen north of the Alps found it difficult enough at first to make a plain bulb and tube and fill it with spirit of wine" (132).

  Roemer proposed that after forming the tube, it be examined for uniformity by examining the length of a drop of mercury as it passed down the bore. If the tube was found to be irregular, it was discarded, and if conical (the length increased or decreased at a constant rate), he took measurements and divided the bore into four equal volumes (67).

  While in many thermometers the bulb was blown on the capillary tube, EB11/Thermometry recommends that it be formed of a separate piece of glass fused onto the stem.

  Bimetallic Thermometers. In Grantville, there should also be thermometers with a dial readout. These have a strip with two different metals layered together, usually brass and iron. The metals have different coefficients of expansion and thus the strip bends toward the less responsive metal. The deflection is proportional to the temperature change and to the square of the length; winding the strip into a helix allows a long and thus more sensitive element to be relatively compact. A pointer is connected to the center. Generally speaking, they are less accurate than liquid expansion thermometers, and require weekly (if not daily) recalibration (Thermoworks), but they are the basis for the most common kind of thermograph.

  Platinum Resistance Thermometers. These, known as RTDs (Resistance Temperature Detectors) rely on the change of electrical resistance with temperature. EB11/Thermometry provides formulae, circuit schematics, and comments on errors and corrections. The current levels must be kept very low (<1 ma) to minimize self-heating (Srivastava 135, 137).

  In the twenty-first century, RTDs are available in two grades, "standard" and "industrial." RTDs will not be found in Grantville homes or schools, but it is conceivable that the power plant has them (most likely "industrial" grade). The standard RTDs are used as primary reference thermometers. They have platinum wire of 99.999% purity wound in a strain-free configuration (MINCO). Unfortunately, the strain-free resistance element is extremely delicate (Ripple), so SPRDs are used in laboratories.

  The industrial grade RTDs use platinum of lower purity and also have a simpler construction in which the resistance element is supported (or thick enough to be self-supporting). When calibrated, they have an accuracy of perhaps 0.01oC, an order of magnitude less than the SPRTDs. But they are also cheaper to make and calibrate (Fluke).

  There is a small quantity of platinum available in Grantville in the form of jewelry, and it may be sufficient for experimentation. Commercial development of RTDs will have to await platinum mining (see Cooper, Mineral Mastery, Grantville Gazette 23) and purification. Developers will have to worry not only about platinum purity, but also about mounting the wire so as to minimize the strain caused by thermal expansion and contraction (Price).

  Ev
en if the wire is not subject to chemical attack, it is mechanically fragile, and the wire is typically protected from the medium by encasing it in a glass, quartz, porcelain, or metal tube (Patranabis 223). A plastic cladding might also work. In any event, the sheathing increases the lag time (Srivastava).

  Platinum's advantages are that it is a noble metal, with a high melting point, and that it has a very linear response over a wide temperature range. Copper is more responsive, and linear over the range -50 to 150oC, but subject to chemical attack. Nickel is even more responsive, and is chemically resistant, but there is no simple formula for calculation of its resistance (MINCO). One can scavenge the nichrome wire heating element from a defunct toaster or heating pad. However, nichrome actually has a rather low temperature sensitivity (Lemieux). My expectation is that the first NTL resistance thermometers will use copper wire.

  Thermistors. In an automated weather station, there's no one to go out and read the mercury (or spirit) level on a conventional thermometer. Hence, some sort of electrically based temperature sensor is needed, and the platinum resistance thermometer (RTD) is too expensive for most meteorological applications.

  A thermistor is a resistor whose resistance is temperature-dependent. In 1833, Faraday discovered that the electrical "resistance of silver sulfide decreased dramatically as temperature increased;" i.e., it is a negative temperature coefficient (NTC) material (Wikipedia). The first commercial thermistor was Ruben's (1930).

  There are thermistors in Grantville; they are the sensing element in the digital clinical thermometer. They are ten times as sensitive as an RTD but their temperature response is highly nonlinear (exponential). Also, a single thermistor has a useful temperature range of not more than 100oC and their maximum temperature of operation is 110oC(Ripple) (so don't take them into the desert). (Industrial RTDs can be used outside the thermistor range.)

  I assume that one of the electrical engineers in Grantville has Dorf's Electrical Engineering Handbook (2d ed., 1997). It discloses that NTC thermistors are "ceramic semiconductors made by sintering mixtures of heavy metal oxides such as manganese, nickel, cobalt, copper and iron" (14). It is known that the automation engineering department in the power plant and public works department have copies of Instrument Engineer’s Handbook Third Edition, edited by Béla Lipták, and The Instrumentation Reference Book, Third Edition, edited by Walt Boyes. Both have extensive sections on thermometry and temperature measurement instrumentation. So that gives us a starting point, but I suspect that these must be purified to very high purity and we must also experiment to find which combinations provide strong temperature dependencies. The simplest type of thermistor to make is probably a bead; the metal oxide powders are combined with a binder (to be determined!) to make a slurry and this is applied to a pair of platinum alloy wires held parallel. The beads are dried and then fired in a furnace at 1100-1400oC to sinter the particles (Lavenuta). Given the infrastructure and experimental requirements, I am doubtful that a practical thermistor can be built before the NTL late 1640s.

  Scale, Range, and Calibration. For the thermometer to be useful in meteorology, we needed to have a way of assuring the comparability of observations made with different thermometers.

  If the scale were an arbitrary one, then the only way of calibrating the scale of a new thermometer would be to place it next to a reference one, expose them to several markedly different temperatures, and then mark the tube of the new one to correspond to the temperatures displayed by the reference one.

  It was realized at a quite early stage that the temperature scale should be defined according to reference points corresponding to readily reproducible laboratory conditions. Then a reference thermometer is not needed at all. By 1702, Roemer proposed a scale in which 7.5 was the melting point of ice and 60 the boiling point of water. A decade later, Fahrenheit experimented with several scales, of which the final one had 32 as the melting point of ice and 96 as human body temperature (now known to be 98.6oF). He extrapolated that on that scale, the boiling point of water would be 212oF, and it was only later that others adopted that as the "hot reference" for his scale (Middleton 78-9). Celsius, in 1742, proposed the melting point of snow as the cold reference and the boiling point of water when air pressure was 25.25 Swedish inches as the hot reference, with 100 degrees in between. Other inventors proposed other scales, and a mid-eighteenth century thermometer featured eighteen scales.

  The modern thermometers found in Grantville are likely to be marked in both Fahrenheit and Celsius, and it is very likely that the scientists and engineers in Grantville will push very hard for one or both of these scales to be universally adopted.

  The portion of the standard temperature scale that is marked on the thermometer is its range. Typically, the bigger the range, the less accurate the reading; for ordinary thermometers, an error equal to 1-2% of the maximum range is not unusual. The outdoor thermometers I own have a range of -50 to +50oC.

  Calibration has three aspects: (1) marking the thermometer scale so as to correspond exactly to the reference scale at the two points and at least roughly at in-between points, (2) tabulating the remaining errors in the thermometer scale, and (3) checking the thermometer from time to time to determine the necessary adjustments for physical changes in the instrument.

  When matter is changing phase (between solid and liquid, or liquid and gas, or solid and gas), as long as both phases are present, the temperature should remain constant. Hence, the melting point of ice and the boiling point of water are, at least theoretically, "fixed points."

  In 1777, the British Royal Society reported on "the best method of adjusting the fixed points of thermometers." They had found that depending on the manufacturer, thermometers could differ by 3.25oF in their measure of the temperature of steam. Accordingly, they gave specific instructions as to the design of the vessel (a cylindrical pot with a cover and a chimney, the latter covered with a loose-fitting tin plate), the placement of the thermometer inside, the application of the heat, and the correction for atmospheric pressure. For the ice point, the Society called for the crushed ice to reach almost to the top of the column, and for provision to made for drainage of the meltwater (Middleton 128).

  he vessel used in the boiling point determination is called a hypsometer, and there is a diagram and brief description in EB11/Thermometry. The boiling point needs to be corrected for differences in pressure from the reference pressure. Characters should not use the correction set forth in EB11, but rather one based on modern steam tables. (The power plant should have them.)

  Some of the precautions recommended by the Society are now known to inhibit superheating, a phenomenon in which liquid water exceeds its nominal boiling point (Chang).

  Modern ice slush and steam calibration baths can achieve accuracies of 0.002oC and 0.1oC respectively (Moore 614).

  Even though we can use the Celsius reference conditions to define a scale from first principles, for meteorological purposes, the range -50 to +50oC is much more useful than one of 0 to 100. For that range, other reference points may prove helpful. (The accuracy that can be expected "without extraordinary attention to purity" is typically about 1oC for most of the transitions (although it is 0.05oC for melting gallium) (Moore). Unfortunately, only a few of the "standard" phase transition baths have temperatures in that range, and we don't have access to gallium (melting point 29.7646oC) or indium (159.5985oC). Mercury is available and melts at -38.8344oC. We might be able to obtain p-xylene (13oC); this would involve isolating it from a natural source (perhaps wood tar) or synthesizing it from readily obtained chemicals. Most syntheses also produce its two isomers, which have different boiling points, and the separation isn't trivial despite that difference.

  Studying physical data on organic compounds (The CRC Handbook should be in Grantville), I have noted some common chemicals with phase transitions in the meteorological temperature range: the boiling points of acetone (56.2oC) and benzene (80.1), and the melting points of tert-butyl alcohol (25.7) or g
lycerol (17.8). In each case, you must be sure that the chemical is pure (so you can rely on the reporting values) and that both phases are present. In general, melting point determinations are better than boiling point ones, because the latter are also affected by atmospheric pressure.

  I have also found reference to the use of crystal transition temperatures, at which a crystalline salt changes form (perhaps as a result of the loss of water of hydration). For example, the transition temperature at which both sodium sulfate decahydrate and anhydrous sodium sulfate coexist is 32.383oC (Middleton 57). Sodium sulfate (Glauber's salt) is commonly used because its transition temperature is close to room temperature and it is easily purified by successive recrystallization. Another possibility, once we have access to chromium ores, is sodium chromate decahydrate, which transitions to the hexahydrate at 19.529oC (Magin; Richards).

  Once the two reference points are marked on the scale, the intermediate points can be marked manually by geometric dividing methods (these are known to the down-timers) or ultimately mechanically by a "dividing engine." Either way, a uniform division is achieved.

  Unfortunately, the behavior of liquid expansion thermometers is not entirely linear. The liquid and glass may change expansion rates with temperature, and the bore might not be uniform.

  Modern precision meteorological thermometers are calibrated by putting the thermometer in an alcohol, water, or paraffin bath that is heated to a series of set temperatures, say 10oC apart (Srivastava 108). Naturally that means that you need a calibrated and even more accurate thermometer for monitoring the bath temperature. The platinum resistance thermometer is excellent for this purpose. (Platinum resistance is highly linear over the meteorological range, Middleton 180) The NWS in 2014 uses an SPRTD (NWSRS 8), but our characters would have to settle for less. Even better, this thermometer is incorporated into a thermostat so that the heat is turned on or shut off as needed to maintain the set point temperature. A table is prepared showing the correction needed by the test thermometer to match the reference thermometer at each of these calibration marks, and the observed applies the correction as appropriate.