Maximum Resolution

Images made through a microscope. All subject types.

Moderators: rjlittlefield, ChrisR, Chris S., Pau

pwnell
Posts: 2029
Joined: Fri Dec 18, 2009 4:59 pm
Location: Tsawwassen, Canada

Maximum Resolution

Post by pwnell »

I have been on a long and futile quest to coach ever more resolution from my BX53 system. Ever since I had my Nikon E200 I wanted more. If I had the money I would get an SEM - I see there are commercial "for home use" units out there but they are still a bit expensive. Therefore my 60x/1.35 has to do.

The most details I could resolve from my test diatom slide up to date is shown below. This is with a 60x/1.35 UPlanSApo objective and an 0.95 NA TLD dry condenser under DIC.

Image

I have just received my TLO (NA 1.40 oil) condenser. I wanted to see if theory and reality really matched one another. I noticed the field was considerably brighter with oil. Contrast was much less than without oil (if I am referring to oil I am specifically talking about oil between the condenser and the underside of the slide. There is always a drop of oil between the objective and the cover slip).

I could perceive a resolution increase through the eye pieces, but it was very small. Naturally I took another photo (stack).

Image

I do not see how I can get anything more out of this microscope system - my understanding is even a 100/1.4 oil objective will not give me any more resolution when coupled with an 18MP APS-C sensor.

What do you think? This is the first time I have ever been able to resolve the pores on the diatom on the right.

EDIT: The spacing between individual pores on the diatom on the right is 250nm. There are 39 "rows" (easier to count the rows than the columns) in a space that I know represents 10µm (using a calibration slide with the same objective/camera). According to my calculations, assuming a wavelength of 550nm, with a 0.9NA condenser and 1.35 objective one can get a maximum resolution of 298nm. With a 1.4 NA condenser and 1.35 objective that figure drops to 244nm. I guess I am pretty close then?

René
Posts: 467
Joined: Sun Oct 08, 2006 4:22 am

Post by René »

This is pretty much what to expect. The 100x has a higher NA and therefore will give better resolution. It also very likely yields better contrast compared to 60X, even with an objective with the same NA. Pores are difficult to image as wall material between pores is very thin, compared with the transverse interstriae, and pore distance is smaller then striae distance. Forget comparison with images on the web, a diatom slide is ´personal´ use only, your specimen might be more fragile, mounting thickness probaly not ideal, making visulaization all the more trickier. DIC systems are generally optimized for either contrast, or resolution, it is a compromise.

Hope that clarifies things, best wishes,
René

Ichthyophthirius
Posts: 1152
Joined: Thu Mar 07, 2013 5:24 am

Post by Ichthyophthirius »

Hi,

The 100x/1.40 will give you higher resolution if used correctly (very thin layer of mounting resin); the APS-C sensor is not limiting (you'd need less than 3MP for that).

For maximum resolution you will have to abandon the DIC system. For the very finest details, DIC gives you less resolution than brightfield microscopy. DIC trades off resolution in favour of contrast. Some companies sell "high resolution" DIC components which increase resolution (lower shear) but even there the problem remains. This is briefly explained here: http://www.nikon.com/products/instrumen ... dex_03.htm

See also here, Figure 10: http://molecular.magnet.fsu.edu/primer/ ... intro.html

The solution: use either brightfield or even better, use oblique illumination, which will give you the highest resolution (even though the images are not as pretty as in DIC).

A 100 year-old 100/1.30 apochromat will essentially give the same resolution in oblique illumination as a modern objective (the condition of the object, like refractive index mismatches and wrong cover glass thickness will be more important than the objective you use).

pwnell
Posts: 2029
Joined: Fri Dec 18, 2009 4:59 pm
Location: Tsawwassen, Canada

Post by pwnell »

Thanks for the info. But why would a 100x/1.40 give me more resolution than a 60x/1.35?

Using R = 1.22λ / (NA(obj) + NA(cond))

I get

R = 1.22(550nm) / (1.35 + 1.40) = 244nm resolution for 60x/1.35 and 1.4 NA condenser.

For the 100x/1.4 I get

R = 1.22(550nm) / (1.40 + 1.40) = 240nm. Surely one cannot really see 4nm increase in resolution?

I'll try oblique / BF. PS: The slide was mounted by Klaus Kemp as a test diatom slide - not sure what mounting resin / method he would have used.

Olympusman
Posts: 5087
Joined: Sun Jan 15, 2012 12:31 pm

Home units SEMs

Post by Olympusman »

A couple of years ago I was invited to a local science center for a demonstration of one of those Toshiba tabletop SEMs and wasn't very impressed. I brought a blood smear slide for my test. It appeared the scans were effective up to about 3000X and then it appeared after that the magnification was empty magnification using digital zoom.
The unit used a low vacuum and did not require sputtering the speciment, which may account for the poor quality.
Michael Reese Much FRMS EMS Bethlehem, Pennsylvania, USA

discomorphella
Posts: 607
Joined: Sun Oct 01, 2006 7:26 pm
Location: NW USA

Post by discomorphella »

DIC resolution is actually fairly complex. There are some papers by Shepard and Wilson et. al. which describe the transfer function for a DIC microscope (e.g. Opt Express. 2008 Nov 24;16(24):19462-79."Partially coherent image formation in differential interference contrast (DIC) microscope"), but they are not what one would call generally accessible. Describing them properly would take a lengthy reply, and it would probably belong in the equipment section. However, there are a few things you could do to peak up your DIC resolution. Using your U-TLO top 1.4 NA top condenser lens will certainly help to increase your resolution for partially coherent brightfield imaging (that's the configuration that your resolution formula is describing). Your lower Wollaston prisms however are probably the series designed for the U-TLD dry top lens, which may not be the optimal lower prisms for the U-TLO. Effectively you could be limiting your objective back focal plane fill (which is the NAcond term in the denominator of your resolution formula, even though, yes, its describing brightfield partially coherent imaging and not DIC). So if you could get ahold of the correct prisms for the U-TLO it might help, although the final diffraction limit is still set by your objective's NA. If you really want to significantly increase your optical scope's resolution, you're better off finding a confocal scope which has improved axial and lateral resolution compared to a conventional setup. First however, you should try to use imageJ or a similar analysis program and see how close to theoretical contrast for a given target pitch you have achieved. When trying to basically hit the diffraction limited resolution for a given optical system, even scant, percent or less levels of aberration can cause measurable degradation in imaging (not that you can see it easily looking at the image with the old mk1 eyeball, you'll need to measure your target contrast numerically). As Rene pointed out, your test target is not ideal, and small changes in the coverglass thickness, RI of the cover and the mounting medium etc will induce some aberrations. For ultrahigh resolution in optical semiconductor diagnostics we actually have custom oil immersion objectives with correction collars to help eliminate exactly those optical path differences. None the less trying to get the most out of your scope is a worthy endeavor (and as I will personally attest to, don't let it take up too much of your total scope time... :) )

David

p.s. according to the U-UCD8 manual, for the various UPlanSApo objectives and upper DIC prism, you will need the following lower prisms depending on your upper slider and top lens etc

Image

edit, if you have the regular slider use this chart

Image

pwnell
Posts: 2029
Joined: Fri Dec 18, 2009 4:59 pm
Location: Tsawwassen, Canada

Post by pwnell »

I knew there was an HC and HR DIC kit (I reviewed both some time ago), but did not realise the condenser wollaston prism were different for TLO. Thanks for the information.

Ichthyophthirius
Posts: 1152
Joined: Thu Mar 07, 2013 5:24 am

Post by Ichthyophthirius »

discomorphella wrote:If you really want to significantly increase your optical scope's resolution, you're better off finding a confocal scope which has improved axial and lateral resolution compared to a conventional setup.
I think this wouldn't be of much help. The diatoms don't fluoresce.

The system would only achieve the full resolution of an NA 1.4 objective if a fluorescent object was embedded in mounting medium matched to the objective (i.e. oil) or placed directly underneath the cover slip.

The diatoms are mounted in a high refractive index (RI) mountant (Naphrax or similar). This is needed to improve the contrast of the preparation. When using a very high, NA 1.4 objective, spherical aberration will start to degrade the image because of this RI mismatch.

I'm wondering if the degradation will be so significant that the image is worse than that of the NA 1.35 objective.

In any case, these diatoms were resolved in the 19th century. It shouldn't be a problem resolving them today, just with a simple 100x NA 1.3 or even 1.25 and oblique illumination.

pwnell
Posts: 2029
Joined: Fri Dec 18, 2009 4:59 pm
Location: Tsawwassen, Canada

Post by pwnell »

Technically, the diatom frustules do not fluoresce, the chloroplast in living diatoms do.

But let's get back to the discussion about resolution using transmitted bright field / DIC. My question was why a 100/1.4 would give me more visible resolution than a 60/1.35, based on my calculations for bright field. If everything is perfect I should get 4nm better resolution, and my question is whether that is significant - it does not seem to be.

discomorphella
Posts: 607
Joined: Sun Oct 01, 2006 7:26 pm
Location: NW USA

Post by discomorphella »

Hi Waldo,
I am going to try and avoid reproducing Born and Wolf "Modern Optics" in this post, but a little math is going to be necessary here...
The formula you are using for specifying resolution is effectively describing the width of the point spread function (PSF) of a simple lens system (a circular aperture with an ideal lens transform function) between its maximum and its first zero crossing (or the distance between the first negative and the first positive crossing, but that's not important).
An infinitely narrow (delta function) object imaged through this system will produce an image that simply reproduces this PSF. So this width is the best you can achieve given the diffraction pattern (PSF) of your system, hence its your diffraction-limited resolution. For two such delta functions, the Rayleigh resolution criterion specifies that you can "resolve" the 2 objects when their respective PSF "images" overlap so that the maximum of one falls on the first zero of the second. Runing through the math, this means that a diffraction limited system has a contrast (intensity difference between the maxima) of ~25% when that overlap occurs. This is the origin of the formula you are using, with a modification for the fact that your objective is being illuminated by light from the condenser, which we can ignore for now. I've attached a slide that shows this. The upper right figure shows 2 such PSF functions, one with 0.25 waves of spherical aberration, the other one ideal. The lower left hand figure illustrates the Rayleigh criterion. The dip in the middle is ~75% of the maxima on either side of it, or conversely, there's 25% contrast between the 2 peaks. If you had 2 ideal objects spaced at 0.61*L/NA where L is your wavelength, that's what you should see.

Image

Clearly there are a large number of factors which affect the imaging system and prevent it from achieving this ideal PSF limit. You can describe any deviations from the ideal imaging condition using a variety of optical models (a good one for microscope objectives that is referenced and used a lot is by Fredrick Lanni and Sara Gibson, J. Opt. Soc. Am. A, 8(10), pp. 1601-1613, Oct 1991 if you are really curious). but basically you can think of any aberrations that degrade your diffraction-limited image as a deviation of the imaging rays from their ideal path. The size of the aberration is then a function of this distance, or more properly the product of the distance and the RI along the path. This is effectively why having a coverslip of a nonstandard thickness, or water instead of oil or RI=1.5 media induce spherical aberrations. Likewise a difference in path from the objective to the occular in a finite conjugte scope. There are many such sources of nonidealities in all of our scopes. A 4nm difference (from NA 1.35 to NA 1.40) in resolution can be inverted to obtain the effective path difference that produced this resolution change, be it from an aberration, or also from an NA difference. As you can imagine, the level of perfection that you'd have to attain in all your optics, including your sample, in order to see this difference in the "dip" contrast is simply unachievable. At least for anyone who is trying to get good pictures, not trying to get a physical review letter on some aspect of diffraction. By the way, as you can see from the upper left figure, one quarter wavelength (L/4) of spherical aberration is sufficient to lower the peak of the PSF to 80% of its value. For your 550 nm light, that's something like 137 nm...not accounting for RI etc...so imagine just what kind of heroic efforts you'd have to go through to shave another 4 nm off that PSF width.
Interestingly enough, a confocal scope has a narrower PSF than a regular BF socpe, so its PSF width is approx. 0.37L/NA to generate that 25% contrast, so for a given target, it will have a 25% "dip" at a smaller distance and hence has better resolving power by this criterion. Note the lower right figure though, no matter how much better the PSF may be, it still falls to zero at 2NA/L since the lens aperture is the limiting factor, and so you will get better contrast on your diatom pores with a confocal, BF or fluorescence, but you can't image what wont make it though the lens, so you can't see pore pitches finer than your NA limit (L/2NA).
Unfortunately, as soon as you are not using good old brightfield and switch to some kind of off-axis illumination (OAI, e.g. oblique) or DIC, the PSF is now different. That NA limit doesn't change though, but your contrast improves for certain targets. So you could get better contrast using a regular widefield scope if you tried some kind of OAI, depending on how the diatom's pores are aligned. But that PSF is for a different post.
Whew. Hope this has not left you in a coma.

David

p.s. this example is from a world apart from biological microscopy, so its using an NA of 2.1 and a wavelength of 1064 nm. The Rayleigh and Strehl ratios are all good though...

pwnell
Posts: 2029
Joined: Fri Dec 18, 2009 4:59 pm
Location: Tsawwassen, Canada

Post by pwnell »

Thanks for your exhaustive reply - much appreciated.

curt0909
Posts: 609
Joined: Wed Oct 26, 2011 10:06 am
Location: Pittsburgh, PA

Post by curt0909 »

I have some first hand experience with differences in DIC shear's effect on resolution and contrast. http://www.youtube.com/watch?v=qvGVoxdy-yM i shot this on a custom DIC set up which gave a very short shear. It was shot with a nikon e plan 1.25 with a 1.25 condenser white light. That particular set up resolved more detail of granules in the WBC compared to my Microphot DIC 60x 1.40 with 1.40 condenser. The Microphot's prism shearing favors contrast over resolution. The custom set up also resolves more detail than my Reichert Diastar DIC 100x 1.30 and 1.40 condenser. The Diastar's shear is somewhere between the microphot and the custom build.

https://www.youtube.com/watch?v=BHfoLFiYkoE Here is common baker's yeast with the nikon planapo 60x 1.40 on a Reichert Diastar microscope. This is much higher resolution compared to the same objective on the nikon microphot. Identical objectives, both with 1.40 condenser NA. Even when using monochromatic light on the microphot the reichert was much higher resolution. Microphot did have better contrast.

http://www.youtube.com/watch?v=QEIAtEfmd_g this was done on the same custom scope with a nikon plan apo 40x 0.95. I didn't have the Microphot at the time but this resolution is much better compared to the microphot.

I've been busy with other projects but I've been planning to document the differences of DIC with identical objectives. I found that using the Reichert combining prism on the Microphot condenser gives my favorite balance of contrast and resolution. Once I do it I'll post the results here.

One other method to increase contrast is using monochromatic light, the lower the frequency the better. Just be careful using blue and UV as they can damage your eyes. Only use this with a camera.

Post Reply Previous topicNext topic