Failure to duplicate Wood's 1909 greenhouse experiment

In 1909 the noted physicist R.W. Wood conducted an experiment aimed at demonstrating that the temperature of a greenhouse was not significantly influenced by the trapping of infrared radiation by its glass windows, but rather was the result of preventing its air from mixing with the cooler air outside the greenhouse. Wood observed less than a degree of difference between two boxes with respectively a glass and a rock salt window, see Wood's article along with comments by W.M. Connolley, to whom I am indebted for drawing Wood's experiment to my attention.

It is customary in physics to confirm experiments by repeating them before accepting them as definitive. Remarkably this custom has not been observed for this experiment, which has simply been accepted unquestioningly for the past hundred years as proof of Wood's suspicion.

We have our doubts about Wood's experiment. First it is documented so sketchily as to make it impossible to duplicate faithfully. Second, it contradicts the understanding of a century of physicists who preceded Wood, starting with Fourier in 1824. Third, calculation of the quantity of heat that should be trapped by the glass window in Wood's experiment shows that fully two-thirds of the heat entering the box fails to be transmitted back through the window, raising the question of how that heat is able to leave the box without raising its temperature significantly. Wood makes no attempt to reconcile his one observation with this elephant in the room.

To get a better understanding of the situation we constructed not two but three boxes differing only in their windows. These are shown below, numbered 1 to 3 from east to west. The location is 37 degrees north of the equator and the photos were taken at about 2 pm on 11/28/09.

View from east.

View from above, with box numbers visible (1 to 3 from left/east to right/west).

View from west.

The boxes are made of corrugated cardboard and measure 12" × 12" × 6". The interior is padded with foam walls spray-painted with Krylon Ultraflat black paint for reasonable blackbody performance. Five sides of the exterior are wrapped with Reynolds aluminum foil to minimize heat transfer by radiation other than through the window in the sixth side. All joints are sealed with Scotch tape to minimize exchange of air between the interior and the exterior. Each box is instrumented with a temperature probe monitoring the air temperature at the center bottom of the box and shielded from the sun with a small parasol. This is in addition to the two probes monitoring each window, described below.

The front opening of each box consists of a 12" × 12" window transparent to visible and near-infrared radiation, thereby allowing more than 85% of the incident sunlight to enter the box while confining the air within the box so as to prevent cooling by mixing with the outside air. Roughly 8% of the 15% not admitted to the box is attributable to reflection at the front and back window surfaces, while the rest consists of near-infrared insolation beyond 2.5 microns in the case of boxes 2 and 3 where the transmittance of glass and acrylic starts to fall off.

Box 1: Saran wrap Premium. This is 0.015 mm thick polyethylene film, which is largely transparent to blackbody radiation at 333 K (60 °C). This serves as a modern-day counterpart to the rock salt window Wood used, polyethylene film not being available in 1909. Both materials serve to assess the effect on temperature of a window that passes infrared radiation while preventing convection to the outside.

Box 2: Window glass, 3/32" thick. This blocks over 90% of blackbody radiation at 333 K. Its thermal conductivity is approximately 1 W/m2/K, or R-1 in SI (European or RSI) units, relevant to temperature drop across the window, while its refractive index is close to 1.54, relevant to reflectivity at the front and back surfaces.

Box 3: Acrylic plate, 3/8" thick. Like glass, this also blocks over 90% of blackbody radiation at 333 K. Its thermal conductivity is approximately .25 W/m2/K, or R-4, while its refractive index is about 1.5.

The windows of boxes 2 and 3 are each instrumented with two K-type thermocouples measuring the temperature at the center of the two surfaces of the window, primarily to monitor the temperature drop across the window. These are shaded with the same parasol used to shield the interior probe, again to avoid spurious heating by direct insolation. It was deemed neither necessary nor practical to instrument Saran wrap in this way, the temperature drop being assumed to be negligible.

Very preliminary tests conducted on November 28, 2009 indicate that boxes 2 and 3 run respectively 15 and 20 degrees hotter than box 1, with a significantly larger drop across box 3's window than box 2's. While this is consistent with the expected differences obtained by theoretical considerations, it is so far removed from what Wood found as to raise the very interesting question of how he was unable to observe any significant difference between his two boxes.

In view of these discrepancies it is remarkable that there has been no previous attempt to duplicate Wood's results, and disconcerting that they have been the basis for a century of claims that infrared trapping plays no significant role in greenhouse warming.

We plan to refine the experiment and these measurements in order to improve both their accuracy and meaning as well as gaining insight into how Wood's observation might have come about, at which time it will be appropriate to release more detailed temperature readings. For example double glazing in the form of a second window of Saran wrap a centimeter from the first may well raise the temperature of box 1 by insulating the window better. Meanwhile suggestions for improving the methodology can be emailed to the author's surname using hostname

Double glazing experiment 12/3/09

We repeated the previous week's experiment with boxes 1 and 3 increased from single glazing to double glazing. The main concern with box 1 had been that its thermal resistance was negligible. We dealt with this by adding a second Saran wrap window with a 6 mm (1/4") air gap between the two windows, a standard for some greenhouses. Air has 40 times the thermal resistance or R-value as glass, and 10 times that of perspex, so this gap would correspond to a ten-inch glass slab. To be sure that there was no convection we divided this thin layer of air into 2" x 2" cells with 1/4 " strips of acetate cut from blank overhead transparency slides.

Each layer of material has two reflective surfaces. Assuming a nominal refractive index of 1.5 for all materials, this entails a reflectivity of 4% at each surface. Two layers of Saran wrap will therefore reflect 16% of the incoming solar radiation, as well as 16% of the outgoing thermal radiation at 20 times the wavelength.

To minimize the differences between the windows we added a second perspex window to box 3, of the same 12" x 12" x 3/8" dimensions as the first, with a negligible air gap between them. Being negligible, there is no additional thermal resistance. Nevertheless the gap is many wavelengths of light whence the reflections from the adjacent surfaces combine incoherently and therefore additively by intensity, for a total of 16% as for the double-glazed Saran wrap window.

At 2 pm on 12/03/09 the box with the double-glazed Saran wrap window reached 65 °C while the box with the two 3/8" perspex sheets reached 80 °C. (We subseqently recorded 82 °C with the latter.) During the previous hour the former climbed rapidly to its limiting temperature while the latter was much slower.

Had we settled for an early measurement we would have concluded that box 1 was the hotter. However box 3 continued to rise and eventually overtook box 1 and went on to become 15 °C higher.

These temperatures were measured about 2/3 of the way to the front (window) side of the box. To explore the variability from front to back we retracted the probes by pulling most of the probe out of the box, taking the precaution to insulate thermally the resulting exposed part of the probe. Box 1 decreased to 62.6 degrees while Box 3 decreased to 76.3 degrees. This decreased the difference between these two boxes from 15 degrees to a little over 13 degrees, still a substantial difference. This is consistent with the hypothesis that these similarly sized boxes exhibited similar convection patterns. We attributed the decrease in this case to the fact that the boxes were aimed upwards towards the sun and that hot air rises, as a simplification of the more complex route no doubt followed by the convection in each box. Had the insolation come from below it is an interesting question whether the window end or the black-bottom end of the box would be hotter.


The conclusions we draw from these experiments are as follows.
  • The greater thermal mass (actual mass times specific heat) of perspex relative to that of air slows down the heating of the former, but is immaterial once thermal equilibrium is reached.
  • The high thermal resistance of the double-glazed saran window of Box 1 did not result in Box 1 heating up the way Box 3 did. Hence relatively little of the incoming insolation energy is leaving via conduction through the 6 mm air gap. Moreover the 2" x 2" cells should prevent any significant convection. This leaves radiation as the only remaining conduit. Box 1 is therefore cool because its window is transparent to infrared radiation.
  • In equilibrium Box 3 is significantly hotter than Box 1, which we attribute to the blocking of infrared radiation by the 3/4" total of perspex, forcing the incoming energy to leave the box by conduction through the perspex (since convection through the perspex is impossible).
  • We view these results as preliminary, and intend to improve on the experimental setup in order to develop a clearer picture of the thermal behavior of the boxes. One particularly important change will be to replace the Saran wrap by a salt window for a closer match to Wood's experiment. In August 2010 we made a start on this but have had to put it aside due to pressure of other work. We hope to be able to resume these experiments in the 2013-2014 time frame. Meanwhile here is a summary of our initial experiments with this setup as performed back then.

    August 2010: Preliminary Results with a Salt Window

    Experimental setup: two boxes with interchangeable quarter-inch-thick windows, one glass, the other salt. As with Wood's setup the boxes were painted black inside and were packed in cotton.

    Each box was equipped with three type TMP05B pulse-width-modulated temperature sensors, two to record the temperatures respectively OVER and UNDER the window (with sensors touching the window in both cases) while the third recorded that of the BOTTOM of the box.

    The six thermometers were daisy-chained and fed into a microprocessor that recorded the temperature of all six thermometers every five seconds.

    To minimize bias from imperfectly balanced boxes and thermometers, when switching windows between the boxes the thermometers stayed with the boxes. That way each box and its three thermometers got to spend equal time with a glass window and a salt window.

    For each of the two windows, glass and salt, the following table lists the average temperatures recorded over and under the window, and at the bottom of the box, for two runs each of 100 seconds (so 20 recordings per run of 6 temperatures per recording), with the windows swapped between runs. Each 100-second run was picked out of a much longer run by waiting until all the temperatures had equilibrated.

    Glass 38.7 °C 54.4 °C 75.4 °C
    Salt 36.5 °C 48.2 °C 74.3 °C

    Take-away points:

  • No such thing as "temperature in the box." In the box with the glass window the temperature inside the box varied by 21 degrees from bottom to top, while the salt window gave an even larger 26 degree variation. In his 1909 paper Wood speaks of "the temperature in the box" which this experiment shows to be a meaningless concept: temperature is highly dependent on exactly where the thermometer is placed in the box.
  • At the bottom of the box the window type makes a difference of only one degree, while at the top (under the window) the difference is more than 6 degrees!
  • The temperature drop across the glass window is nearly 16 degrees while that across the salt window is less than 12 degrees. If one pictures each window as two resistors in parallel, one for (thermal) conduction and the other for radiation, the observed drops are consistent with both windows having similar thermal conductances but with the salt window's radiation resistance being much lower.
  • Ideally swapping the windows should have made no difference, but there was sufficient variation between runs to indicate badly manufactured boxes. Although switching the windows should compensate for most of that imbalance I would like to remake the boxes more carefully for better balance. These results are therefore only preliminary, and I plan to find time to rebuild this setup, hopefully in the next few months if possible since people have been bugging me about this lately.

    Vaughan Pratt
    Professor Emeritus
    Stanford University
    Stanford, CA 94305-9045