20241111M update - 

20xExM Single-shot 20-fold expansion microscopy
https://www.nature.com/articles/s41592-024-02454-9

==>The ideal microscope for imaging 20xExM is likely to be FISHscope (with GPU deconvolution) since this research grade widefield fluorescence microscope is much faster imagng than confocal microscopes. Our current high resolution objective lens is our 60x/1.35NA, 108nm XY pixel size, recommend Z-step 300nm, working distance 300 um (so maximum 15um "original size" if use full 20x expansion). Working distance of most Evident Scientific (ex-Olympus) high resolution (X-Line and HR) objective lenses have smaller working distances (i.e. 1.45 and 1.50 objective lenses) and the 20x vs (say) 10x expansion doubles resolution anyway. A couple of potential tweaks to image better and brighter and less photobleaching (many also mentioned elsewhere in this and other McTips -- and often applicable tostandard fluorescence microscopy):

• use 1.5H high precision 170 um +/- 5 um coverglasses -- maybe even hand measure each coverglass and use those that are 170um (+/- say 1um) for imaging.
• If need BOTH 20x expansion AND further than the 300um working distance, evaluate #00 (~70um, if can find commercially) or #0 (~100um) or #1 (~130um) to gain ~100um or ~70um or ~40um "real world" Z range (I'm ignoring the refractive index of your mounting medium in these values -- Z-distances are with respect to air). #1.5H and #0 coverglass imaging dishes are available from Mattek. If using R.I. unmatched mounting media, will have spherical aberration that may (hope) or not (bummer) be corrected by the cellSens software GPU deconvolution. I note that Expansion is tunable, so could expand say 15x to stay within working distance (if specimen is in imaging dish, could do Z-series, evaluate, adjust R.I. and expansion amount, re-acquire).
• Refractive index match your mounting medium to 1.518, match borosilicate coverglass and the immersion oil. As close as possible is good (and knowing the R.I. helps some deconvolution algorithms). 
• D2O does not quench orange, red and near infrared fluorophores, whereas H2O does (H2O is a lousy quencher, but at 55 Molar ~100 -OH groups close to a fluorophore in solution, H2O quenches most of these dyes). So: Consider D2O ("deuterium water") as the starting point for your mounting medium, ideally adjust R.I. to 1.518 -- see Maillard 2020 ChemSci see http://confocal.jhu.edu/mctips/af610-x  (Alexa Fluor 610 was the brightest fluorophore in D2O or H2O of 42 fluorophores the authors tested). 

p.r.
https://www.sciencedaily.com/releases/2024/10/241011141242.htm

With 20-fold expansion, researchers can get down to a resolution of about 20 nanometers, using a conventional light microscope. This allows them see cell structures like microtubules and mitochondria, as well as clusters of proteins.

In the new study, the researchers set out to perform 20-fold expansion with only a single step. This meant that they had to find a gel that was both extremely absorbent and mechanically stable, so that it wouldn't fall apart when expanded 20-fold.
Using this technique, the researchers were able to image many tiny structures within brain cells, including structures called synaptic nanocolumns. These are clusters of proteins that are arranged in a specific way at neuronal synapses, allowing neurons to communicate with each other via secretion of neurotransmitters such as dopamine.

20241108F - FISHscope cell details 20x 0.75NA vs 60x 1.35NA raw vs GPU deconvolution - details can matter.

Note: no images in this section - the acquired images are for the user and their lab, not for my essay.

Yesterday I trained a new user on FISHscope http://confocal.jhu.edu/current-equipment/fishscope

Description      Mag       N.A.   working distance (mm)    Pixel size GM Recommend Z-step

UPL SAPO      20x        0.75 air      0.6                                 324 nm             1,000 nm or 500nm
UPL SAPO      60x        1.35 oil       0.3                                108 nm                300 nm or 200nm (or 400nm)

Sub-confluent thin cells (~4 um thickness?) growing as well separated colonies on a coverglass. The background fluorescence was much higher in the dense middle of large coonies compared to colony edges and small colonies. 

Several observations on my part:

* unknown coverglass thckness - for specimens with potential high resokution imaging, buy and use #1.5H (H = High Precision, 170 +/- 5 um) thickness coverglass (and double check that 2 coverglasses are not stuck together) -- see box below (20241014M).

* huge variation in colony sizes: this type of variability potentially affects both image quality and the biology they are studying. 

* 60x oil lens, 25% 395nm LED power: DAPI not only photobleached but photonverted to green fluorescence - nuclei became bright green in the "Alexa Fluor 488" channel. Mitigation:

• I changed to 10% LED power (I recommend 10% as the lowest power on the lumencor lamp, because may not be stable at less than 10%).
• Unknown mounting medium: (i) every user doing imagign should know EXACTLY what is on their specimen, (ii) I recommended testing THermoFisher Prolong Glass (no DAPI), with volatile compound being allowed to outgas for at least 72 hours in a large volume of air (ex: back of closed drawer, NOT tiny sealed slide box).

* We initially trained on 20x, single plane. Not much subcellular detail.

* 60x single plane raw images were more "somewhat more" informative.

* 60x Z-series raw data (22 or 30 planes, 300nm Z-spacing) better, but blurry (because 60x/1.35NA lens has ~600nm in-focus, but if cell is 3000nm thick, then ~2400nm "not quite" focused).

* 60x Z-series GPU deconvolution (Process menu -> Deconvolution -> Constrained Iterative, 25 iterations) was was much more informative than 20x single plane or 60x raw. I note that the "C.I." algorithm on FISHscope requires Z-series, though maybe not as thick as we did (though maybe would have been even better if we did say 12um instead of 9um = 9000 nm of 30 planes 300nm Z-steps).

Key item: the deconvolved Z-series best focus planes clearly showed sub-cellular immunofluorescence details, such as endosomes and endoplasmic reticulum, not readily recognized in raw data. 

* Junk on coverglass: there were a lot of small "objects" (puncta?) on the coverglass - readily apparent in the GPU deconvolved Z-series (Z-panes at the coverglass). I suspect "antibody aggregates" and recommend:

• use high quality antibodies (year old is likely bad - but even one month in a refrigerator might ruin one or more antibodies).
• use mini-centrifuge ("minifuge") in a cold (4 C) rotor to spin down any aggregates (and take just the supernatant, not resuspend the pellet) -- balance the minifuge. See below for reference(s) (seach this web page text for centrifuge). 

* huge variation in colony sizes - a decade ago at UMiami I coauthored a Blood paper with X. Jiang, I.Lossos and colleagues (2010, PubMed 20844236) - see panel E and F of https://pubmed.ncbi.nlm.nih.gov/20844236/#&gid=article-figures&pid=figure-2-uid-1 for HeLa "tetrads" of 4 grand-daughters 40 hours after plating dilute suspension; fit very nicely in 63x objective lens 2x confocal microscope zoom -- helps that HeL are not particularly motile; we also did 3-plane stacks, 0um, 2um and 4um from the coverglass-media surface, and reflection interference contrast microscopy to foucs consistenttly at the coverglass-media interface). If more cells are needed, users could optimize trypsinization, cell density, culture time, for their needs (ex: 60 hours would probably result in 8-cell HeLa colonies). Optimizing cell density at plating would also provide good density of colonies, not too sparse (ineffieinct searching) and not too dense overlap).

* only two slides imaged ... you and your P.I. decide on what controls to do. If weird stuff appears in your images, if you run the appropriate controls every time, you can troubleshoot (provided you use 60x oil Z-series deconvolution when looking for subcellular details). If you do not have the controls, your choice, your thinking. 

----------

Unknown: would there have been more information -- compared to the 60x/1.35NA lens -- if 

a) FISHscope has a 100x/1.4 NA objective lens (~60 nm pixel size) or 150x/1.45 NA (43 nm pixel size)?

b) FISHscope wdetecting direct labeled antibody using BV421 (emission peak ~420nm) or BUV395 (emission peak ~395nm) instead of Alexa Fluor 488 (emission peak ~520nm)?

c) confocal with the same lenses (i.e. move specimen to our Olympus FV3000RS confocal with same lenses)?

Note: 108nm is FISHscope pixel size with the 60x/1.35NA objective lens using our Hamamatsu ORCA-FLASH4.0LT with 6.5x6.5um pixels (please donate to us the Hamamatsu Quest2 back-illuminated quantitative CMOS camera, 4.6x4.6um pixel size, peak 95% quantum efficiency s peak ~85% for 4.0LT) dxy is the XY resolution, My rule of thumb is dxy / 3 = optimum pixel size (though dxy / 3.5 might provide more information) -- basically Nyquist sampling theorem (2.2 data points for a sine wave) for 2D biological specimens. I also note deconvolution improves resolution by ~10% (if specimen, coverglass #1.5H, bright labeling, excellent camera, excellent optics and calculations are perfect, and no photobleaching).

To make absolutely clear: We prioritized 60x/1.35 NA objective lens for FISHscope for single molecule RNA FISH: 108x108nm pixel size results in more area (~11,664 um^2) compared to 100x lens (65x65um pixels = 4225 um^2, 36%(calculations assume 100% fill factor for the pixels - probably not the case). For smFISH, pohotons matter more than resolution.

dxy widefield                       = 0.61 * wavelength / NA

dxy confocal 1.0 Airy unit)  = 0.51 * wavelength / NA

dxy confocal 0.66 Airy units = 0.8 * 0.51 * wavelength / NA ... 0.51 * 0.8 = 0.41.

dxy (WF) = 0.61 * 500 nm / 1.35 = 226nm ... deconv (0.9) = 203nm ...  108nm pixel size too big on purpose.

dxy (WF) = 0.61 * 500 nm / 1.40 = 217nm ... deconv (0.9) = 196nm ... ~60nm pixel size good match (3x is 180nm).

dxy (WF) = 0.61 * 500 nm / 1.45 = 210nm ... deconv (0.9) = 189  nm ... 189/3.5 = 43nm oixel size, maybe more info.

skipping 1.0 Airy unit and just calculating 0.66 Airy units ("confocal sweet spot" re Jeff Reece, NIH, sweet range).

dxy (0.66AU) = 0.41 * 500 nm / 1.35 = 152nm ... deconv (0.9) = 137nm

dxy (0.66AU) = 0.41 * 500 nm / 1.40 = 146nm ... deconv (0.9) = 132nm

dxy (0.66AU) = 0.41 * 500 nm / 1.45 = 141nm ... deconv (0.9) = 127nm

 what if use wavelength 395nm (BUV395, ignoring light source issue wrt confocal):

dxy (WF)       = 0.61 * 395 nm / 1.45 = 166nm ... deconv (0.9) = 150nm

dxy (0.66AU) = 0.41 * 395 nm / 1.45 = 111nm ... deconv (0.9) = 101nm

I note that the last value, 101nm is better than commercial SIM (structured illumination microscope) systems, which is ~2x better resolution than standard 10 Airy unit confocal (but Alexa Fluor 488 etc -- SIM systems may or not be capable of exvitation of BUV395 and would not have BUV filter set standard).

20241014M    Coverglass info:

No. 00           ~70 um    (0.07 mm)

No. 0            ~100 um   (0.08 - 0.12mm)
No. 1            ~145 um   (0.13 - 0.16mm)
No. 1.5         ~170 um   (0.16 - 0.19mm)

No. 1.5H        170 +/- 5 um (High precision) Ex: Marienfeld (Germany).
No. 2             ~210 um  (0.19 - 0.23mm)

notes:

* I strongly recommend buying and routinely using #1.5H coverglass

* #00 ... I communicated with Marienfeld (10/2024) ... #00 orders are doable, minimum 1000 pieces ... in the U.S., order through one of their two U.S. distributors (see their distributor wb page for contact info). 

* Sometimes 2 coverglasses can be stuck together when you take them out of the box. I encourage checking EVERY coverglass for "pairs". A pair can be recognized by Newton's Rings (interference fringes, similar to a thin layer of oil on a water puddle) appearing when you hold the coverglass(es) up to a light (ceiling light works fine). Positive control: you can hold 2 together. Another method is inspecting the edge: 2 stuck together appears thicker. You can also inspect the coverglass on edge (10x objective lens has a 1+ mm field of view; or use a mono-zoom or stereo microscope).

* wrong coverglass thickness can result in spherical aberrations.

* GM: if you refractive index match the immersion media (i.e. oil), coverglass (R.I. 1.518), and mounting media (i.e. cured Prolong Glass, R.I. 1.518) the light rays travel in a straight line = best imaging quality. This implies you could use a #00 or #0 or #1 coverglass and get additional working distance to see deeper into your specimen. 

20241014M   

Amount of antibody in a typical immunofluorescence experiment or other antibody experiment

page 111 (pdf page 11):
4. Calculate the molar ratio between antibody and DNA.
(a) The concentration of the antibody after the concentration step is ~4 mg/mL.
The molecular weight of a secondary antibody is ~150 kDa = 150*10^3 g/mol.
The molar concentration of the antibody is therefore [see text for equation] = 26 67 μM. at 4 mg/mL
                                                                                                                          26.67 nM at 4 ug/mL                                                                                                                                                                7      nM at 1 ng/mL

Typical antibody is 100 ug in 100 uL, so 1 ug/ul = 1 mg/mL.

I note that a Nanobody (VHH-only heavy chain, usually from alpaca or llama, aka "camelids") is ~15 kDa, so all concentrations would be 10x higher. See Nano-Tag.com and Chromotek - now https://www.ptglab.com/ (in U.S. distributed through ThermoFisher) for commercial fluorescent dye labeled secondary antibodies (both sourced from pleiner et al 2018 J Cell Biol). 

Florian Schueder, Ralf Jungmann (2024) In Situ Imaging of Proteins Using DNA-PAINT Super-Resolution Microscopy in Christoph Wulfing and Robert F. Murphy (eds.), Imaging Cell Signaling, Methods in Molecular Biology, vol. 2800: 103-113 (Chp9). https://doi.org/10.1007/978-1-0716-3834-7_9

20241011F GM favorite tools (mostly software)

X1.com ... Windows PC indexing program.

NCH Debut ... screen and audio recording software (see OBS Studio to have multiple webcams on screen). 

Drive cloning softweare - examples

• MiniTools Partition Wizard ... clone PC drive(s) [need destination drive, adapter] (JHU I.T. does not want this on medical Domain PCs and may band from non-domain PCs). 
• NTI Echo 6 ... alternative drive cloning tool to MiniTools ... NTI offers bundle with NVMe SSD enclosure and USB cable. 
• Note: I've had enough bad experiences with Acronis to NOT recommend them. 
• JHU I.T. controlled computers (computers on 'medical domain'): You may need to have I.T. clone your computer drive onto a larger capacity drive. 

SSD drives: NVMe M.2 drives 1TB (TeraByte) capacity and largers are usually (PCIe gen3, x4) ~3 GB/sec (3 GigaByte/sec ~ 3000 MegaByte/sec) vs 256 GB and 512 GB NVMe typically only 600 GB/sec, same as mSATA and SATA-6 SSDs. 

Motherboard performance (x16 slot) ... so NVMe drive x4 is 1/4 speed for x16 slot - can get Add-In Card (AIC) that can house 4 (or more) NVMe drives for one x16 slot (see RAID info below):

• PCIe gen3 x16 motherboard slot ~15 GB/sec. ... most PC's in past decade (i.e. circa 2014 - 2024)
• PCIe gen4 x16 motherboard slot ~31 GB/sec. ... some PC's past ~4 years (circa 2020 - 2024)
• PCIe gen5 x16 motherboard slot ~61 GB/sec. ... newest (and most expensive) past year or so (2024)
• See below for DDR6 and previousgeneration RAM speeds.

Resizable BAR = Modern (circa 2020 plus - usually later gen3 and newer genX) motherboards feature to turn on.

• "Resizable BAR (Base Address Register) is a PCIe capability. This is a mechanism that allows the PCIe device, such as a discrete graphics card, to negotiate the BAR size to optimize system resources. Enabling this functionality can result in a performance improvement."
• Before this, NVidia GPUs were limited to 256 MB (MegaByte) data transfers. 
• AMD (the other major GPU vendor) called theirs SAM = "Smart Access Memory", see https://www.howtogeek.com/819578/what-is-resizable-bar-on-a-gpu for more information. 

GPU (NVidia GPU cards dominates scientific software such as spatial deconvolution are mostly written only for NVidia CUDA driver interface) (Graphical RAM speed typically increase per generation):

TITAN (circa 2023):                        ~3Teraflop (FP32) (probably PCIe first gen)

RTX 2080 (circa 2016);                  "few Teraflops" (10?) PCIe gen2?

RTX 3090 (circa 2019);                  ~20 Teraflops  ... PCIe x16 gen3 

RTX 4090 (circa 2022);                  ~90 Teraflops ... PCIe x16 gen4 

RTX 5090 (expected Feb 2025); ~120 Teraflops ... PCIe x16 gen5 

GM suggestion for NVidia and AMD: since top ofthe line GPUs are 2 or 3 PCIe slots wide to accomodate chillers and fans, add  second PCIe x16 gen5 interface to DOUBLE or TRIPLE the data transfer rate. Sure, more complexity, so what? 

NVMe drives and arrays ... current Highpoint RAID PCIe gen5 AIC (add-in card) ~61 GigaBytes / second. Example:

• $1500 HighPoint Technologies Rocket 1608A PCIe Gen5 x16 to 8-M.2x4 NVMe Switch AIC  {up to 8 drives, need four or more for full speed}
•  URL https://www.amazon.com/HighPoint-Technologies-Rocket-8-M-2x4-Switch/dp/B0D328GPLB 
•  Delivers 64GB/s of Bandwidth & Real-world Sustained transfer speeds up to 56,000MB/s
• amazon ads of previous generation Highpoint drives stated that using two AIC cards would double the speeds. GM hypothesizes this is not mentioned for the 1608A AIC card because the limit may be PC motherboard and RAM speeds {gm 20241020S - added DDR6 ram speed below, "DDR6 up to 134.4 GB/s+", the "+" is potential for overclocking -- upshot is DDR6 RAM is ~2x single Highpoint (or similar) NVMe PCIe x16 gen5 card, so RAID'ing two might increase performance (and 3 AICs might increase a little bit more) -- payoff of 2 or 3 AIC cards is price of AIC cards + NVMe drives could enable a "virtual drive" of massive size at a much lower total cost than real DDR6 RAM. Virtual drives are still possible on Windows 10 and11
  • If you try it, and the software (web browser interface) enables configuring across AIC cards, congratulations and please let me know (gmcnamara@jhmi.edu and geomcnamara@earthlink.net).
  • [Windows 10 Pro] Control Panel -> Systems -> Advanced system settings [bottom of page] -> "System properties" dialog box -> Advanced Tab -> Performance section -> Settings ... button -> Advanced tab -> Virtual memory -> Change... button -- text description: A paging file is an area on the hard disk that Windowsuses as if it were RAM". GM note: be sure drive array is configured optimally - and do not be surprised if paging results in faster 'wear and tear' on the NVMe drives (turn on SMART monitoring and replacedrives before they fail).
    Also: standard Windows 10/11 for home andProfessional have limited maximum RAM access -- see Windows web site or discuss ith your I.T. dept about Windows Server or other Windows versions that can access 1TB and above RAM and virtual RAM. ... Also: what applications are you running that needs massive amount of RAM?
• This model has cooling fan and warning lights - likely also software warning. 
• Example 1TB NVMe gen5 drive $171 each
• https://www.amazon.com/SABRENT-Advanced-Performance-Internal-SB-RKT5-1TB/dp/B0CXV7MG6Q
• four 1TB drives      $684
• eight 1TB drives $1,368
• Example 2TB drive $306 each
• https://www.amazon.com/SABRENT-Advanced-Performance-Internal-SB-RKT5-1TB/dp/B0CXVGC466 
• eight 2TB drives $2,448
• Note: RAID0 is "striped" (fastest, no redundancy), RAID1 is mirrored, RAID5 is classic redundancy (ex. 3  of 4 -- or 7 of 8 -- drives  used as data space, one drive for "parity")
• Note: the motherboard, te RAID card and all the drives need to be PCIe gen5 to go full speed. If any component is older, you are limited in speed by that component. "Gen5" was introduced around 2024, so any older motherboards are slower. 
• By coincidence 20241011F I helped a faculty member set up a Highpoint RAID AIC card. Was easy - web browser interface for installation AFTER running "Disk Managment" to initialize their four NVMe drives as "GPT".
• NVMe speeds, single drives:
• 1TB and up (approximate): 3-3.5 GB/sec for gen3, 6-7 GB/sec for gen4, 12-14 GB/sec for gen5.
• 256 and 512 GB: often limited to ~0.6 GB/sec (600 MB/sec) similar to SATA-6 drives.

***

[this sub-section added 20241020S[

RAM speeds - faster is better, and more expensive ... and need newer motherboard etc

Current (2024) DDR6 PC ram speeds in Gigabytes/sec - for comparison with PCIe x16 slot speeds, NVMe drive arrays, GPU transfer speeds (and resizable BAR memory transfer aperture)

https://www.pcworld.com/article/2237799/ddr6-ram-what-you-should-already-know-about-the-upcoming-ram-standard.html

• In a cross-generational comparison of memory bandwidth, DDR6 would once again increase significantly:
• DDR up to 3.2 GB/s
• DDR2 up to 8.5 GB/s
• DDR3 up to 17.0 GB/s
• DDR4 up to 28.8 GB/s
• DDR5 up to 67.2 GB/s
• DDR6 up to 134.4 GB/s+
• As things stand, it can therefore be assumed that the fastest DDR6 memory modules will be able to provide at least 134.4GB/s of memory bandwidth, with OC modules delivering significantly more memory throughput per second.

***

Hamrick.com VueScan Pro - universal scanning software for flatbed scanners and some 35mm film scanners.

Flatbed scanner(s) - ideally with transparency lid.

35 mm film scanner(s) - with adapter(s) for microscope slides, maximum area is 36x24mm, examples:

• Meyer Instruments - Pathscan Enabler ... current model Pathscan Enabler V (fifth model). My original Tiki_Goddess scan was on original Enabler -- https://works.bepress.com/gmcnamara for multiple Tiki_Goddess images 
• Plustek 35 mm film scanner + additional "film strip" holders - I adapted one for microscope slide double sided tape to bridge across opening). Pixel size is ~4.5um "at specimen" so similar to a "typical" 4x or 5x objective lens - but scanner has a larger field of view.

CrystalDiskMark 8 and CrystalDiskInfo (8) ... a bit tricky finding the "download installer" links (buttons) on the web site

•   main site https://crystalmark.info/en
•   I suggest buttons immediately below "Quick"  at https://crystalmark.info/en/   (takes several seconds for file downloads to appear to start).
•    Note: the "Mark" benchmark speeds may be much greater than Windows 10/11 file transfer or programs ("Apps") speeds. By going with a fast NVMe drive or RAID array, the hardware is not the bottleneck. 

OBS Studio ... enables simultaneous use of multiple web cameras ... GM recommends using different model for each web cam to help keep organized.

Glary Disk Cleaner ... weekly purge junk files from Windows PCs - avoid installing "Accessories" (junk).

Gary Undelete ... best to install before you need to recover any files!  - avoid installing "Accessories" (junk).

PC speakers - I now recommend buy with AC power, generally enables louder output than USB powered. For work, no woofer (bass); for home but one with woofer (bass). 

Uni-Ball 0.7mm black pens; red pens.

Sharpies - "magic markers" - available in many colors and formats.

Tools (literally):

• Jewelers screwdrivers set (extremely useful for fine work)
• Diamond scribe (fine glass cutter) - scratching coverglass facilitates precise break (super sharp glass edge)
• Levels (long i.e. 48" and medium length i.e. 36")
• Tape Measure (25 foot probably long enough to purchase, can always place a mark and measure more)
• Allen wrench set metric
• Allen wrench set "English"
• Box cutters = exacto knife (made infamous by 9/11 hijackers - do not bring on plane)
• Pliers (many sizes, styles), such as needle nose, wire cutting, etc
• adjustable wrenches (can buy set or sets)
• "wrench set" with tool pouch" (I think  of these as "scrscent wrenches", ex. https://www.amazon.com/SWANLAKE-32-Piece-Combination-7mm-22mm-Vanadium/dp/B0BQLLH659 
• Big wrench(es) - often needed for leverage (torque) to attach/detach air regulator on gas tank (100% air, 5% CO2, etc). ... i.e. 3 foot long
• "wrench extender tool bar" - usable with Allen wrenches, crescent wrenches etc - apply greater torque
• screwdrivers "slot"
• screwdricvers "Phillips"
• Drill (battery powered may be handy), large set of drill bits, drill saw bits, sanding bits. Interesting confocal listserv thread 10/2024 mentioned making your own plastic imaging dishes by (1) operate drill bit "in reverse" to gently make hole and then move around to enlarge hole, (2) sanding bit(s) to smooth surfaces.
• Files (large and small sets) to file down and smooth edges (see also sanding drill bits above). 
• Contractor-Grade Alignment Pump Wedges ... (amazon) AirShim Inflatable Pry Bars and Leveling Tools (was unafamiliar with these until a Hudson Robotics engineer made use of two of these 10/2024 - extremely useful to raise Biosorter up to remove Vistek vibration isolaton pads prior to temporary move).
• [none of my cores have 10/2024, look very useful) - tool cabinet(s) on wheels, to organize tool set(s). Some drawers could also store optical components, fiber optics, liquid light guide.
• Label maker - ex. Brother PT-### (can buy cheaper 3rd party replacement cartridges). Label everything! for example, both end of computer cables AND where they go on computers and equipment. 
• Vibration isolation pads ( for under compressor, etc) - can buy rubber pads from amazon.
• ==> Likely a lot of types of tools I am missing from this list - you can look through amazon.com and tool companies web sites for more ... can often order with next day delivery if missing stuff.

Yes, 63x /1.4NA objective lens will typically be more useful than a 100x/1.4 NA objective lens, for the larger field of view advantage you mentioned. This ASSUMES they are each in “perfect” condition. If either has been old and abused, the better quality one will be superior.
The LSM880 is based on a laser scanning microscope , so the combination of Zoom and Number of pixels (in each of XY) determines field of view (zoom) and pixel size. Please DO NOT TRUST the settings recommended by the vendors!!!
(new): Z-step size and doing “AiryScan deconvolution” (880 might offer “double deconvolution) matters. More on this below.
Scan speed matters … faster and more averaging better (if no photobleaching). … if AiryScan offers “photon counting” then use “line accumulation” to add the counts (I do not believe AiryScan has this).
Wavelengths matter (“Lambda”) … confocal and AiryScan resolution has the wavelength (in practice, the emission wavelength) in the resolution equations. So: Brilliant Violet 421 (BV421), laser line 405nm,  emission max 421 nm (I would set emission wavelength range to something like 420-450nm would give better resolution than Alexa Fluor 488 (488ex, 500-540em), Alexa Fluor 647 (640ex, 655-695em), etc. If purchase BV421, do not use DAPI or Hoechst DNA counterstain (in fact, best to not use any counterstain).  … BV421 antibodies and streptavidin are available from BD Biosciences and BioLegend.
Probably not practical for your experiment: the best resolution would be 405nm laser line (shortest wavelength available on confocals/AiryScans), “reflection mode” AiryScan, as in Sivaguru 2019 (PDF attached).  Note: the microscope companies do not always do a great job with reflection mode and it could turn out that 405nm laser line is not good with Reflection mode. Also need to understand that the coverglass has relatively strong reflections (can be useful, but not in your case), which can be mitigated by perfectly matching the refractive index of the coverglass (oil 1.518 is standard, mounting medium Prolong Glass after 3+ days to solidify should also be RI ~1.518). The PDF gives a couple of ways of generating reflection labels – colloidal gold is a classic approach; NanoProbes.com offers “NanoGold” (may or not be detectable clearly on AiryScan, so may need one of their Amplification methods).
I am assuming you are doing immunofluorescence labeling. Brighter fluorophores (or more of them in a small volume) enables better, faster imaging. Andrew Belmont (https://pubmed.ncbi.nlm.nih.gov/30154186/  see also pubmed search Belmont AS tyramide) reported tyramide signal amplification (TSA) generates ~50nm diffusion radius around antigen:1stAntibody:HRP(s) – and can be decreased by “adding stuff” to the reaction buffer (ex. Sucrose to increase viscosity, so could be say 25nm radius, smaller than your target 100nm).
The objective lenses are designed for 170um thick coverglass. This is classically #1.5 coverglass, but the variability in manufacturing is suck that you should use #1.5H = High Precision coverglasses. These are mostly manufactured by Marienfeld (Germany), and at least one format is available for sale on the Zeiss web store. If using imaging dishes, Mattek (-0.170-) and iibidi offer #1.5H coverglass (price is about 2x the #1.5 dishes .. a trivial additional cost compared to say 1 hour of confocal microscope time in a core).. Quick tip: after you buy 1.5H coverglasses, segregate all other coverglasses to a drawer labeled “use only if doing low resolution microscopy”. Since any experiment might work well enough to be moved to AiryScan, probably best to standardize on #1.5H.
FYI – Aleksandr Smirnov, Neuroscience Core (PCTB 10th floor office) manages an Abberior STED nanoscope, with ideal specimens, Aleksandr told me 20nm XY resolution, ~60nm Z-resolution (with Abberior deconvolution … and STED typically good for one focus plane before everything photobleaches … I suspect optimum scan speed (fast with photon counting accumulation), better fluorophores, mounting media, careful choice of laser power, could enable Z-series). So: if you need to validate your 100nm resolution, contact Aleksandr for STED time.

Confocal resolution … then AiryScan (divide confocal 1.0AU by 2)  then pixel size (divide AiryScan in this example by 3.5:

Dxy (1.0 Airy Unit)= 0.51 * Lambda / NA

    Alexa Fluor 488 would then be  0.51 * 520nm / 1.4 = 189nm … so AiryScan divide by 2 = 90nm … pixel size   25.7nm … I would round down to 25nm.

    Brilliant Violet 421                       0.51 * 440nm / 1.4 = 160nm … so AiryScan divide by 2 = 80nm … pixel size    22  nm … I would round down to 20nm.

Z-step size: the Z-resolution equation is complicated enough for me to avoid memorizing it, so I simplify that for high NA objective lenses,  and skipping some steps, I would use

Z-step(AiryScan) as 3x the XY pixel size, so:

Alexa Fluor 488            … 75nm.

Brilliant Violet BV421 … 60nm.

Note: the best resolution for AiryScan requires doing the post-processing step(s) – I don’t know if this is “double deconvolution” or some other term.  Also, do not assume Zeiss has the best math! You might want to get a demo license (2 weeks on one of your lab computers) of Huygens, see https://svi.nl/Huygens-Array-Detector-Software   which may compute better.

**

If you are going to do AiryScan “at the limit’ a lot, you may want to contact our local Zeiss rep to get a demo of a new, higher NA, objective lens (and find the money to buy it if it works great). Zeiss offers 1.45NA objective lenses – usually for TIRF, but with perfect refractive index matching can acquire beautiful confocal or AiryScan, every 0.05 NA improvement increases resolution (if perfect specimen, coverglass etc). These lenses have limited working distance (say 100 um vs 200um for 1.40 NA lenses), but you want to image near the coverglass (under 20um) anyway.

àlastly, quick trick: if Zeiss has limited line averaging, can do line and frame averaging … can also do smaller Z-step size (say 2x as many) which the deconvolution software will more or less treat as equivalent to extra averaging.