When combing optical stimulation and electrophysiology for the first time, you may be disappointed to learn that the large voltage swings you record in response to light presentation are not necessarily the result of a physiological process. Instead, they may be caused by a direct interaction between the light and the electrodes, independent of the effects of any light-gated ion channels.
These light artifacts are often referred to as the "photoelectric effect," but this is a misnomer. The photoelectric effect occurs when electrons are emitted from a metallic surface in response to light, and usually requires wavelengths in the ultraviolet range. The light artifacts seen during optogenetics experiments are more likely to result from the Becquerel effect, caused by the unequal illumination of electrodes placed in an electrolyte. The effect is named for Henri Becquerel, the famous French physicist.
What to look for
Light artifacts can either show up in continuous (local field potential) recordings or sampled (spike) recordings. In the local field potential, they usually appear as an exponential charging and discharging, with a duration of approximately twice the length of the light stimulus. For a 2 ms stimulus, then, you would expect to see a light artifact of approximately 4 ms. If the light intensity is very high, the discharge phase of the light artifact can last longer (see figure to the right). Still, as far as the LFP is concerned, the best way to distinguish artifact from genuine light activation is to deliver light pulses that are an order of magnitude shorter than the physiological effects you expect to see. For example, the EPSPs or IPSPs caused by light-induced action potentials will last for tens of milliseconds, even if the original light pulses last a few milliseconds.
In spike recordings, the initial transient caused by the light artifact may be above threshold, and will trigger the acquisition of a spike. Care must be taken to examine waveforms the co-occur with the beginning and end of light pulses, to make sure they look like actual spikes. Any artifactual "spikes" will be very tightly time-locked to the light pulse, whereas real induced spikes will have more temporal jitter.
Importantly, at least for input-inverted recordings with nichrome wire electrodes, the light artifact is always positive-going (same direction as extracellular spike peaks). This makes it easier to distinguish the artifact from physiological events with the opposite polarity. [TODO: Has anyone seen different effects with another type of electrode?]
How to prevent it
Light artifacts result from direct exposure of the recording electrode to light. Electrodes closer to the light source, such as those glued directly to a fiber, are much more susceptible to light artifacts. Therefore, the easiest way to prevent light artifact is to move electrodes away from the light source (see figure to the right), or to shield them from the light with an opaque surface, such as painted glass. Reducing the pulse width can also greatly diminish the artifact.
Fibers mounted directly above the recording sites of silicon probes should be a concern, although artifacts are not always seen in practice. If artifacts are a problem in this situation, it may be possible to illuminate the tissue from another angle, in order to reduce the light intensity to which the recording surface is directly exposed.
A much more involved—but ultimately more effective—solution is to change the composition of your electrodes. Some metals, such as indium tin oxide, are resistant to the Becquerel effect. Efforts are under way to develop electrodes coated with such metals to eliminate light artifacts (see Zorzos, Dietrich, Franzesi et al., 2009 Society for Neuroscience Abstracts). Dipping nichrome wire in indium tin oxide nanoparticles, followed by high-temperature sintering, results in greatly attenuated light artifacts. Alternatively, electrodes can be fabricated directly from indium-tin-oxide-coated substrates, yielding artifact-proof linear electrode arrays.