"Our glucose sensor is very sensitive and responds to concentrations that are inaccessible to conventional glucose monitoring technologies," team leader AT Charlie Johnson told nanotechweb.org. "These low concentrations fall within the range at which glucose is found in saliva."

The researchers began by fabricating thousands of pairs of metal electrodes, each separated by 10 μm, on a silicon wafer. Next, they deposited a commercially available solution of semiconducting nanotubes (from NanoIntegris Inc) across the wafer. This resulted in nanotube networks bridging the gaps between the electrodes, so making thousands of nearly identical, independent electronic circuits.

"We then bound a compound called pyrene boronic acid to the sidewall of the nanotubes," explained Johnson. "This is a bifunctional, highly negatively charged molecule that strongly absorbs onto a nanotube at the pyrene end and covalently binds to glucose at the boronic acid end."

When the nanotube circuit is placed in a solution containing glucose, the glucose molecules diffuse to the sensor and form a pair of covalent bonds with the boronic acid. The acid impedes the conduction of charge carriers (electrons and holes) that are moving through the CNT circuit underneath – in a phenomenon called carrier scattering.

"As a consequence, the resistance of the circuit increases and the measured current through the device drops," said Johnson. "This drop in current is our sensing signal and a larger concentration of glucose induces a large drop in current as more boronate anions are formed. By looking at the percentage drop, we are thus able to infer the exact concentration of glucose in the solution being tested."

Most commercially available glucose sensors on the market today detect glucose levels in blood, which means that the patient has to go through the ordeal of pricking their finger several times a day and dabbing the sensor with blood. Accuracy is sometimes a problem with such sensors, too, because readings can be sensitive to temperature, humidity and other interferents in blood – such as hemotocrit and vitamin C. What is more, these devices are able to detect glucose down to limits of just 100 μM, which limits them to analysing blood samples only.

"Our nanotube sensor is much more sensitive – we measured an absolute limit of 300 nM. This means that other body fluids, such as saliva, which contain much less glucose, can be tested," says Johnson. "Using the sensor we describe, typical glucose saliva concentrations (of between 40–80 μm) are well within the detectable range."

The researchers say that there is already "considerable" commercial interest in the new sensor, but that they still need to "demonstrate the efficacy" of the device in saliva before such devices appear on the market.

To this end, the Penn team is now busy testing samples of saliva in the lab and comparing the results with analyses on clean water. "Ultimately, we would like to compare saliva samples from healthy people with those from people suffering from diabetes," revealed Johnson. "We believe our device could greatly improve diabetic patients' quality of life while still maintaining a high level of diagnostic accuracy."

The present results can be consulted for free on the arXiv server and have been published in Applied Physics Letters.