Does Thermal Radiation Travel From Cool To Hot Bodies

The Background

There is a belief that the cold atmosphere with CO2 (and other Green House Gases - GHGs) cannot keep the earth warmer than if there were no GHGs. A good source of these comments are WUWT
Michael Moon says: February 6, 2013 at 1:11 pm
Silver Ralph,
You just flunked your first hourly in Thermo. The cooler radiator would be warmed by the warmer radiator, and begin radiating more. If you think this would warm the warmer radiator, then you will fail all your hourlies and never get through school.
Wilis, Joe Public has it exactly right. If you want to know what happens to the flux from a cooler source when it hits a warmer source, the answer is exactly nothing. It is not absorbed, but immediately re-emitted, transferring NO heat.
All these analogies are amusing but ignore Second Law.
The sticking point is usually verbalised as "you cannot warm a hot object with a cool object".  And this is true, but you can slow down the cooling of a hot object by placing a cooler object (that is warmer than the background) next to it.

Radiation travels in straight lines – generally! – and when it leaves the cool object it does not know whether it will hit a hotter or cooler object. It is only the albedo that will determine how much  radiation is reflected or absorbed by the objecty it hits and albedo changes little with temperature in the range considered in climatology.

In the extremes:
If it is a mirror finish in wavelengths considered then no radiation will be absorbed (and since emissivity usually parallels albedo then no radiation will be emitted);
If it is a true black body at the wavelength considered then all radiation hitting it from whatever source will be absorbed (and similary emitted - according to its temperature).

Note that a body may reflect at one wavelength but absorb at another. This effect is not considered here.

Radiation passes both ways from the hot body to the cool AND from the cool body to the hot, the additional energy from the cool body will reduce the net energy flow from the hotter and the hot body will therefore cool more slowly. The radiation from the hot body will reduce the net energy flow from the cool but additionally may actually put more radiation into the cool than is being emitted, thus heating the cool body.
From the above it stands to reason that if both bodies have unchanging internal or external sources of energy  then the hot body temperature will reach a final temperature higher than if the cool body were not present.
Oxygen and Nitrogen molecules do not significantly absorb or emit LW radiation.

Note that the vertical scale is logarithmic each division represents a 10x increase in effect
An atmosphere without GHGs will not impede incoming or outgoing radiation significantly. The body of the earth will receive full solar radiation and it surface would heat up to a temperature that will, as a near black body, radiate the same quantity of radiation directly to space at a temperature of  2.7K (i.e. -270°C).

In addition to the solar radiation the body of the earth would receive radiation from the 2.7K temperature of space.

GHGs absorb some  of the long wave infra red radiation from the body of the earth. These will then almost instantly emitted in all directions. After multiple absorptions and re-emissions there will be effectively 50% up and 50% down. The downward radiation ( “back radiation”) will be from an atmosphere much warmer than the 2.7K of space. The total radiation hitting the earth will be solar+back radiation+2.7K from space. With a warm GHG atmosphere the net radiation is still to space but the body of the earth is receiving addition energy from the GHGs and so will be warmer.

Without GHGs there is very little radiation from the rest of the atmosphere - O2 N2 etc absorb and emit little radiative energy (they will of course conduct and convect). But a gas atmosphere without GHGs will not radiate to space. The body of the earth will of course cool since this will STILL radiate - its radiation passing straight through the O2 N2 molecules.
This experiment attempts prove or disprove the supposition that a warm object cools more slowly in the presence of a cool object (warmer than background but cooler than the hot object).

The test setup:

The sidewalls are approximately 150mm high and the whole unit rests on a wooden bench.

The double layer cling film barrier attempts to stabilise the environment around the hot plate when the cool side changes from a warm object to background condition. It is intended to isolate the hot plate from conduction and convection on the cool side whilst passing IR radiation.

The two plates were heated using a hot air gun. The hot plate is heated to well above the maximum range limit for the camera (120°C). The temperature is then allowed to drop to 120°C whilst the environment stabilised

The temperature difference between hot and warm plates was approximately 18°C at the start of measurement. At 120°C the camera records a sequence at 6.5 frames per second with maximum noise reduction (ie the cameras internal algorithm averages out noise on  the measurements). A reasonably high data rate is chosen to allow averaging to be applied to the data if necessary. 

The picture shows the setup with hot and warm plates (18°C temperature difference) facing each other through a “double glazed” convection and conduction barrier (the vertical black line on the left). The temperature shown between plates is the wall temperature not the air temperature (see earlier post showing that air and water vapour are not picked up on the thermal camera http://climateandstuff.blogspot.co.uk/2012/12/water-vapour-and-thermal-imaging.html)

The wall temperature on the hot side is 43.9°C – this is the temperature of the wall approximately 250 seconds after heating was stopped.

This picture shows the no warm plate scenario approximately 230 seconds from heating having stopped. Note that the wall temperature in the hot compartment - 44.7°C is similar to that in the hot/warm setup - 43.9°C.  The open side of the setup is facing a matt black surface at approximately 21°C. The data from the run is extracted using suitably placed measurement areas on the 2 plates:

Only area AR02 is of interest – the hot plate temperature – although any area can be analysed after recording.

The temperature of each pixel in the defined area (6x19 pixels approximately) is averaged.

The two runs provide series of data with the hot plate cooling from approximately 120°C to 50°C with temperatures measured every 154ms.

The rate of cooling should be the same when cooling from the same temperature with or without the warm plate if the statement "You can slow down the cooling of a hot object by placing a cooler object (that is warmer than the background) next to it" is not true.

The two curves have to be aligned such that at one time period the two temperatures are identical. This is manually adjusted in the spreadsheet (available on request). 120.0431°C (with warm plate) and 120.0784°C (no warm plate) are the closest match temperatures These are at a time of approximately 1 minute from start of recoding.

A second plot has been produced that is sychronised at 67°C. This shows that the result is not just anomaly caused by the heating process

A total of 5700 results per run were obtained and analysed

The Results:

synchronised to 120°C

Synchronised to 67°C


From this it seems clear that when a hot and warm plate are interacting radiatively, the hot plate cools significantly slower than if there is no warm plate.

If the initial data up to 600seconds is ignored there is still an increasing deviation in the plot showing the plate is not cooling as quickly.
Obviously the hot plate does not get hotter - it still cools, but at a slower rate!

Does the thermal barrier pass Infra Red?

The photo below shows the with and without temperature of a hand measured with the camera (it should be noted that the camera response does not extend over the whole IR range – only 2µm to 13µm). The temperature difference shows that 2 layers of "cling film" absorb a significant amount of IR but not ALL of it (hand temperature with clingfilm is 30.2°C and without 35°C ) In order to bring the temperatures of the hand back to 35°C the emissivity correction in the camera has to be changed from 0.92 to 0.58.

Notes: Sources of error Cooling of hotplate will be affected by:
  • Conducted and convective heat loss. These have been minimised by the barrier and the 4 sided corregated card enclosure.
  • Room temperature variation Temperature was measured at 21C before and after each run.
  • Air currents. Minimised by the enclosure and the bench location
  • Radiation from hot surfaces (including the cling film window). This should be the same for both runs (note tat these temperatures will be less than the hot plate so could be considered to be part of the experiment.
Inaccuracies - Most are neglegible:
  • The camera self calibrates before the start of each run.
  • Internal temperature of camera changes during run - because this is a continuous recording the usual timed self calibration is stopped. It is possible that drift may occur. This was minimised by allowing 60 minutes warm up and a fixed ambient temperature. Also both runs would suffer from similar drift.
  • The absolute temperature accuracy will not affect results as the same temperature range is measured
Confirmation bias:
I  obviously would like to show proof of the statements made in the introduction, however the results are extracted and plotted by a simple spread sheet which can be made available, There is no manual operation at this stage.
Calculations are simple and carried out on both sets of data identically

All I can say is that these are the results obtained from this simple experiment without modification.

Please criticise and suggest improvements either email or comment (be_very_careful[at)hotmaildotcom)

Additional stuff.

The results seemed too good to be true so I extracted the data from slightly different areas and redid the spreadsheet calculations.

Note different shape of plate sense areas and AR03 sensing back wall temp on hot side

Synchronising to 120deg C at time 0 Absolute temperature  - note backwall temperatures now included

Synchronised to 120C diffenece between hot plate and warm plate and hot only setups

Resynchronising at 76degC at time zero Absolute temperature

Still get widening temperature gap showing coolingof a hot plate is slowed by a warm plate

So again these 2 runs analysed differently show that a cool object (warmer than background) can slow the cooling of a hot plate


Thoughts for a repeat:
Continuous monitoring of ambient temperature.
Enclose hot plate in box of constructed with thermal insulation with 2 sides of cling film (limits heat loss to mainly IR).
Put constant power into the hot plate (fixed voltage across a resistor) Measure temperature with thermocouple - set to approximately 80C when facing ambient.
With hot plate facing plate atambient temperature
Allow temperature to stabilise
Remove ambient plate (possibly now above ambient)
Replace with warm plate at 70C initially
Allow temperature to "stabilise" should rise then fall as the warm plate cools
Remove warm plate leaving side open to ambient. 

Allow temperature to stabilise
Possible problem is the not quite thermally transparent cling film. - as this heats it will begin radiating. 

Improvement suggestions?


  1. What happens if the cool plate is at room temperature?

  2. Same temperature as background so it would follow the cooling curve of the no warm plate scenario.

    With the earth you are looking at 2.7K bacgrond temp of space vs the radiation from GHGs

  3. I like the idea of your attempting to demonstrate a rather difficult concept with a simple experiment, but the only thing I think you've shown here is that the "warm" plate has affected the environment of the "hot" plate so that it loses heat more slowly.
    I think that probably two identical boxes, one for each plate, would be required, and that your experiment would have to be repeated a number of times with varying plate temperatures (even "cold" plates) and varying the distances between the boxes. Also, your experimental results should be repeatable by interchanging hot and warm plates from box to box.

  4. Hot plate side is same construction with and without warm plate. Temperature at start of analysis of hot plate is same with and without warm plate. Convection and conduction from the hot plate should therefore be the "same" with and without the warm plate.

    The similarity of hot side conditions is shown in the second analysis where the backwall temperature is measured. from 76degC the temperatures of the back wall is about 1degC

    The warm plate is isolated from the hot plate area by a double glazed window of "cling film". The cardboard walls joining the compartments will transmit very little of the heat. The walls of the compartments are tall to minimise the convection from the warm side changing the convection of the hot side.

    The main changes between the 2 runs is the radiation from the warm plate passing through the double glazed window.

    The Hot plate cannot be swapped to the warm plate side as the conditions are different - without the warm plate the side wall is removed to expose the double glazed window to ambient temperature 20C approx black body at 0.3m distant.

    I fully agree that the experiment needs to be repeated, but I have to get 3 hours of time and the thermal camera! I have an idea to repeat the test using finned heatsinks which if the correct type should "focus" the IR in line with the fins, I think I will also run the camera at 60Hz frame rate to allow a better smoothing and perhaps just record the hot plate (giving a greater area for the camera to average). Don't know when I will get the time and camera simultaneously. It would be better if this were repeated by others

  5. Impressive work. Have you repeated the experiment already? Are you aware of anybody else having conducted such?

  6. Another set of experiments are here.

    Still have to do the "ultimate!" experiment but need a controlled environment (better than+-0.1C)

  7. Nice set up! One of the arguments is that a warmer object can not add "back radiation" from a relatively cooler emitting body because the higher energy, orbital states of the atoms of the warmer body are already filled. That is, the warmer body is constantly emitting photons, whereas any incoming, lower energy photons from the cooler body can not be received, and will simply be reflected and disbursed.

    How about this scenario: As the warmer body does emit photons, incoming photons from back radiation may immediately fill those vacant states? Even though they are lower energy, ie lower frequency, they might be finding perfect orbital states during the outgoing flux of photons. They would simply be competing at the surface of radiative emission with other photons within the black body that are also trying to get to the surface. Basically, the warmer black body surface doesn't give a hoot where the photons come from. (Scientifically speaking :) )