Depends wether you want to couple into a fiber later on. Fiber-Fiber couplers have potentially little loss.

It also depends a lot on the mirrors you use when doing free space. Depending on wether you use Al or Ag mirrors you loose around 1-10% per mirror.

So if you have multiple cheap mirrors in your 10m beam path, the optical fiber would be the better solution.

Another consideration might me laser safety, fiber is usually safer in this regard.

Depends on how you define "better". its not a quantitative term. For example, since c=c0/n, light travels faster in vacuum than glass (n~1.5). In terms of loss, free space may be fine in the lab, but obstructed by rain or snow in the real world, in addition to divergence loss. You need to be more specific.

Also, if you are delivering ultrashort pulses, dispersion in the delivery system needs to be in consideration too. Both fibers and free space optics has dispersion factors related to their optical elements. For fibers, ultrashort pulses also experience dispersion due to nonlinear responses.

Even without rain or snow... dust can be an issue!

In Canada, there's only rain or snow because no one smokes anymore :)

Perhaps tobacco smoking is less, but I have witnessed the rise of marijuana. :)

Think about this. Suppose that I want connect to points in the earth, but these point are separated for mountains or what else so these point do not see each other. In this case, it is impossible the direct connection. You have to include several communication stations in order to avoid all the obstacles.

If you use optical fibers, you will not have these problems.

Well, You can use ionsphere as a waveguide as people have done it during the past decades. But I have no idea how much is the loss for ionsphere light transportation with rspect to a fiber-optic communication. Besides, the refractive index for the upper atmosphere is not the same as that for the surface anymore.

Actually that is depends on your applications. If you want pass the light intrinsically then you can use the Optical fiber or if you want pass the light extrinsically you have to use the space...

Long haul, metro- and the last mile, i.e. fiber-to-the-home have proved that the fiber is viable application, at least...We are all stay connected on this web page thanks to the undersea-laid optical fiber. Yes, this is a different world of industrial applications, efficiency of delivery from point A to point B is respectable with typical 0.2 dB/km as propagation losses. On top of this you have bending losses, coupling losses etc. The point is that you have to go to all these nitty gritty details to get a complete cost-benefit analysis for the intended application. You have to be very specific otherwise you'd generate the endless discussion with answers that may not be relevant for your purposes...

For example: since you've tagged the question with "Optics" and "AFM" I can guess that you may want to couple some laser (Ti:sapph or cw?) to a near-field scanning optical microscope (NSOM) and since two units are in separate rooms you're contemplating an idea of either drilling a wall and going free-space or laying down a fiber...

As many of the previous answers it depends on your application. if you are looking for 10 m it depends on the requirements of your system in terms of data-rate, sensitivity of the receiver, indoor or outdoor environments, line of sight or not and obviously BER.

Therefore initially you should define your system as indoor/outdoor and data/bit rate. FSO systems are a good choice for medium long distances ( 10 m - 10 km) on very stable outdoor system with LOS, where attenuation is not a big deal and the data rate is not every high ( 1Gbps). optical fibre transmission is more efficient in terms of attenuation and data rates at any distance compare with a classic FSO system, however the deployment of those systems are further more expensive than an FSO system.

In the meantime I advise you to define more clearly the requirements of your system not just the range.

As previous answers show, I need to define more accurate my problem. Actually, I am working on an optical fiber based system that acts as a small AFM. In fact, transmitters and receivers are fibers (all other stuff are like astandard AFM). Our preliminary results show that free space setup works more efficient in terms of signal to noise ratio and being easy to handle, so, I was suspect about the fiber itself. Because for 10 m light pass, it seems the total loss should be negligible. But still compared to free space, the efficiency of fiber system is lower.

As usual the answer is, that it depends. The instrinsic loss of the fiber highly depends on the wavelength and fiber type and can be as low as 0.2 dB/km (no, that's not a type it's dB/km) for single mode 9um core telecom grade fiber at wavelengths around 1550nm. So if you don't use exotic fiber or wavelengths the actual fiber loss should be quite low, however and here is the caveat, you will have to couple the light into the fiber first. And again the coupling efficiency depends on the fiber type and beam profile. A good gaussian beam can be coupled to a single mode fiber (9um core diameter) with an efficiency of approx. 1dB. The stability of this coupling against environmental changes highly depends on the design of your coupling and the stability of the laaser beam and might have a negative impact on the overall SNR. So without knowing the details and requirements of your setup you can't really tell what is "better". You should first describe in more detail how exactly the setup looks like. Also I didn't really understand why you need to send the light over a distance of 10m is the light is already in a fiber, or did I get that wrong?

Agree with answers above, in case of 10m it completely depends on details of your application. If you can feel free to run the laser beam over 10m (say, environment and safety regulations allow), what is the reason to introduce a fiber and accordingly additional optics and insertion losses?

There has been some work with putting laser transceivers on the tops of buildings in dense urban areas. They are reasonably reliable since they have line-of-sight with each other and are traveling a relatively short distance via multiple hops from building to building. When you go any significant distance though, weather and dust starts to become a potential issue.

But you have not explained us what you want connect. Depend of each case, the answer could be different.

I echo what others have said: It depends on what you mean by "better".  The same is true for the term "more efficient".  For example efficient in terms of power use or efficient in terms of capital investment?

Setting aside your 10m example, free space optics (FSO) are generally useful from a few tens of meters up to about 5 km in ground based terrestrial systems.  The higher you get in the atmosphere the better so mountain top to mountain top might be a few tens of kilometers and in space ...

There are two advantages of ground-based FSO that I know.  First, FSO latency is better than fiber-optics.  Second, FSO is better in CapEx than fiber optics.  FSO frequencies are not regulated and (line-of-sight) does not require digging up roads or leasing another companies dark fiber or buying their service.  For last-mile, short range backhaul and low latency, FSO wins (unless the fiber is already there and you own it).

In communications, the real comparison is between FSO and microwave.  Microwave also suffers from atmospheric effects (different from FSO due to frequencies) but over about 70 years of development has created an acceptable offering in communications.  This is effected by proper link planning and the same can be done with FSO.