You're Welcome! On the air-solid interface, thermal drift is usually an issue when you are close to the resolution limit or atomical resolution of crystal lattices (or when performing STM). In other words, when system and sample have reached thermal equilibrium it is somewhere in the few nanometers-range per hour. The largest drift artifacts are usually due to your sample. For example we have several MultiMode devices in our group that offer a very low intrinsic drift. When I started my work I used a bad double sided adhesive tape to fix my samples and got drifts in the range of several nanometers per minute making images without drift correction useless. ;-) 

In force spectroscopy thermal drift is a huge issue. Due to their advantages we regularily use the Olympus Biolever Probes. When we immerse them into solutions or change temperature, usually we need to wait for half an hour. The gold coated cantilevers are actually bimetals and therefore strongly bending due to temperature changes. When everything is in thermal equilibrium we are usually able to measure until our biological samples "die" or the solvent is evaporated (a big issue when using CHCl3). Again instrumental drift is not a big issue compared to the factors above. More important is the fluid cell design (open/closed, size, possibility to change temperature) and - especially for biological samples - is it possibly to setup the fluid cell externally and then transfer it to the SFM? Most fluid cells are build up in the SFM (for example by approaching the cantilever in air to the surface and filling them with solution afterwards). 

Yet again, if you are going to use somehow structured surfaces you may check the xy-long time position stability and reproducibility of tip positioning of the SFM device.

The height and force resolution of state-of-the-art AFM devices is far below levels you achieve in fluid cell imaging. Those values are usually obtained on atomically flat surfaces on the air-surface interface resulting in something around 50 pm RMS noise. Fluid damping, thermal drift, sample adhesion to the surface and many other factors are of much larger importance in solution imaging. For example, you may recieve very low RMS noise values just by decreasing bandwith on any device. But in solution you won't be able to measure anything using such parameters.  ;-) Force resolution for force spectroscopy is anyway limited by the cantilever rather than the instrumental noise. 

Usually SFM distributors give you the opportunity to test your specific samples at their laboratories/headquarters. Afterwards you have a good idea, which device performs best there and not on single-crystal surfaces. An important thing to take in mind is also instrumental setup. Some systems are "made for" fluid measurements, for others this is just "an option". I personally like those devices that can be mounted on inverted optically (fluorescent) microscopes and measure quasi "in-petri-dish" (such as JPK NanoWizard 3, Oxford MFP-3D, Bruker Bioscope Catalyst or AIST CombiScope). Also the improved "force volume"-modes are great for fluid imaging (Quantitative Imaging, Force Mapping Mode or PeakForce). Finally you may also take a look on long-term support and upgrade policy of the company producing your SFM of choice. 

Thanks Manuel  for your  useful insight. Could you comment of the thermal stability over a period of several hrs with a state of the art AFM?

You're Welcome! On the air-solid interface, thermal drift is usually an issue when you are close to the resolution limit or atomical resolution of crystal lattices (or when performing STM). In other words, when system and sample have reached thermal equilibrium it is somewhere in the few nanometers-range per hour. The largest drift artifacts are usually due to your sample. For example we have several MultiMode devices in our group that offer a very low intrinsic drift. When I started my work I used a bad double sided adhesive tape to fix my samples and got drifts in the range of several nanometers per minute making images without drift correction useless. ;-) 

In force spectroscopy thermal drift is a huge issue. Due to their advantages we regularily use the Olympus Biolever Probes. When we immerse them into solutions or change temperature, usually we need to wait for half an hour. The gold coated cantilevers are actually bimetals and therefore strongly bending due to temperature changes. When everything is in thermal equilibrium we are usually able to measure until our biological samples "die" or the solvent is evaporated (a big issue when using CHCl3). Again instrumental drift is not a big issue compared to the factors above. More important is the fluid cell design (open/closed, size, possibility to change temperature) and - especially for biological samples - is it possibly to setup the fluid cell externally and then transfer it to the SFM? Most fluid cells are build up in the SFM (for example by approaching the cantilever in air to the surface and filling them with solution afterwards). 

Yet again, if you are going to use somehow structured surfaces you may check the xy-long time position stability and reproducibility of tip positioning of the SFM device.

You might also want to check systems driving the cantilever vibration optically. I recently purchased a Cypher ES with the Bludrive option (it's a product from Asylum/Oxford Instruments). It is not as flexible as the systems mentioned above in terms of sample size and accessibility (e.g. unsuitable for large samples like petri dishes and no way of fitting an inverted microscope) but it has a good temperature control and the ability to exchange fluids from a sealed cell while engaged.

As pointed out by Manuel, the actual noise/resolution limit is unlikely to be determined by the machine sensitivity for advanced AFMs, but more by your choice of cantilever, thermal stability and your particular setup/experiment. My priority is high-resolution and stability, and the Cypher ES excels on both counts. This is not some kind of advertising and I know different users prefer different types of machines, but if stability is a concern, I strongly suggest you look for AFMs with optical oscillation of the cantilver. In my case, I was able to follow atomic details of a surfaces in water for > 5 hours without any external intervention on the image parameters (setpoint, frequency, drive etc). Obviously the system was allowed to thermalize first, but In my experience, this is not possible with conventional acoustic drive due to a time evolution of the system's transfer function.

That's  really good.  Atomic details of a surfaces in water for > 5 hours without any external intervention on the image parameters (setpoint, frequency, drive etc)!