To find out what molecules are present in a material we use spectroscopy. It goes back to Newton and his observation that a beam of white light passing through a quarz crystal, or through a glass, gets refracted into many colours. In better physical terms, each type of material has a unique refracting index that affects in different ways light wavelengths causing the appearance of colours (that is a separation of wavelengths, each one perceived as a colour by our eyes).
In the following centuries this uniqueness has started to be exploited to identify the material composition and more and more refined techniques have been discovered to get more and more precise and sensitive results. This latter refers to the quantity of material needed to appreciate its impact on the light and hence to be able to identify it. We can use just traces of a substance and stil be able to detect it (since it is Summer in the North Hemisphere and vacation time you may want to read some of Deaver thrillers with detective Rhyme a wizard in getting to the bad guys by using spectroscopy). However, these "traces" contains billions of molecules.
More sophisticated methods (like gas chromatography and Raman spectroscopy) can detect single molecules provided they are sufficiently big, that is contains a few hundred atoms (a protein contains thousands of atoms, so does a DNA gene...).
Now researchers at Rice University have come up with a new sensor that can detect molecules as small as 20 atoms. To do this they use a variant of the Raman spectroscopy, a technique based on the fact that when a shower of photons hits a molecules some of the photons (very very few, most pass through the molecules unscathed, and when I say "most" I am talking about a ratio that is 1 photon interacting out of a thousand billions that pass through) interact with the molecules and change their energy level in a very specific way, so specific that can identify the molecule.
Since so few photons actually change their energy level the signal is very very weak and undetectable unless it gets amplified. Normal amplifier used in Raman spectroscopy manage an amplification between 100 million to 10 billion times, amazing but not enough!
This is what the researchers at Rice managed to do: increase the amplification by using two lasers that along with 4 gold discs reach an amplification of 100 billion times.
We are clearly at the edges of sophistication in the sensor area but it is good to see the kind of progress that are being made. And I personally feel that the improvement in sensors (in sensitivity, power consumption and cost) will lead to a completely new way of understanding and interacting with the environment and with the ambient we live in. Our work at EIT ICT Labs on Smart Spaces has just started to scratch the amazing opportunities ahead.