Protocol for DRIFT-EM: Direct wafer retrieval of ultrathin serial sections for volume electron microscopy

Summary

Large-volume serial electron microscopy enables reconstructions of neural circuits at synaptic resolution but depends on reliable collection of large libraries of ultrathin sections. Here, we present a low-cost protocol for direct wafer-based section collection. We describe steps for modifying a commercial diamond knife, constructing thermoplastic barriers, mounting and tuning static ionizers for section clearing, and automated region of interest mapping of collected sections.

For complete details on the use and execution of this protocol, please refer to Medina and Gogola et al.¹

Graphical abstract

Highlights

• •Steps for constructing a modified diamond knife and wafer boat for on-wafer collection
• •Procedures for thermoplastic-barrier application and ionizer placement
• •Guidance on section collection and wafer drying
• •Instructions for ROI detection, mapping, and SEM imaging and processing

Publisher’s note: Undertaking any experimental protocol requires adherence to local institutional guidelines for laboratory safety and ethics.

Large-volume serial electron microscopy enables reconstructions of neural circuits at synaptic resolution but depends on reliable collection of large libraries of ultrathin sections. Here, we present a low-cost protocol for direct wafer-based section collection. We describe steps for modifying a commercial diamond knife, constructing thermoplastic barriers, mounting and tuning static ionizers for section clearing, and automated region of interest mapping of collected sections.

Before you begin

The protocol below describes the specific steps for collecting ultrathin tissue sections onto 100 mm silicon wafers for scanning electron microscopy (SEM). These steps assume tissue has been chemically fixed, heavy-metal stained, dehydrated, and embedded in epoxy resin (Epon or LX-112) according to standard connectomics sample preparation protocols² though we also provide a brief protocol for this preparation. The DRIFT-EM system consists of three components fabricated or assembled in advance: (1) a modified diamond knife with a 3D-printed wafer attachment, (2) silicon wafers fitted with hydrophobic thermoplastic barriers, and (3) static ionizers mounted for positioning near the ultramicrotome. All three must be prepared before proceeding to the step-by-step protocol, and are described in detail below.

Innovation

DRIFT-EM advances direct wafer-based ultrathin section collection for vEM by replacing expensive magnetic or pneumatic clearing systems with inexpensive, off-the-shelf static ionizers. Key innovations include: (1) a 3D-printed wafer attachment that fits a standard commercial diamond knife and holds a 100 mm silicon wafer level during sectioning; (2) hydrophobic thermoplastic (ethylene-vinyl acetate; EVA) or 3D-printed acrylonitrile styrene acrylate (ASA) barriers that physically confine and do not stick to floating sections over the wafer surface, eliminating the need for magnetic resin impregnation; and (3) static ionizers that generate a gentle, laminar electrohydrodynamic (EHD) flow that clears newly cut sections away from the knife edge without disturbing the water meniscus. Together these elements enable continuous collection of 2,000–3,000 sections at 40 nm thickness for approximately $1,000–2,000 USD in total setup cost - two orders of magnitude less than commercial tape-based systems^3,4,5 - while achieving section packing densities comparable to state-of-the-art magnetic approaches.^6,7

Institutional permissions

All procedures involving vertebrate animals must be performed in accordance with relevant institutional and national guidelines and regulations. The experiments described in this protocol were performed under University of Chicago IACUC-approved protocol #72480. Researchers implementing this protocol must obtain equivalent approvals from their own institutional animal care committees before proceeding.

Preparation 1: Fabricate and assemble the modified knife boat

Timing: ∼1–2 days (machining + glue cure)

In this preparation, a commercial diamond knife is machined to accept the custom 3D-printed wafer attachment, and the attachment is permanently bonded to the knife boat. All CAD files (STL) for 3D-printed components are available at: https://github.com/gwildenberg/DRIFT-EM-CAD-files/tree/main.

• 1.Print the wafer boat attachment (Figure 1A), and optional ASA barrier ring using polyactic acid (PLA) or ASA filament on an FDM (fused deposition modeling) 3D printer (e.g., Bambu Lab X1C).

Note: Use 0.12 mm layer height; optimize speed, infill, and support settings to minimize surface roughness and ensure watertightness of the wafer holder.

Note: STL files are available at the GitHub link above and at Zenodo: https://doi.org/10.5281/zenodo.19380076.

Note: PLA is suitable for the wafer boat; ASA is preferred for the barrier ring due to its natural tendency to repel floating ultrathin sections, thus aiding in keeping ultrathin sections away from the edge of the silicon wafer and not sticking to sections when water is withdrawn.

• 2.
  Machine a commercial Diatome Ultra-Maxi diamond knife (Figure 1B) using a 2-flute carbide endmill.

  • a.
    Mill away the aluminum side walls and the back portion of the knife boat at a ∼6° descending angle relative to the diamond blade until a flat boat surface is achieved.

Note: Do not remove aluminum around the set screw and avoid getting too close to the diamond knife on the front side of the aluminum boat (i.e., the side where the knife is attached).

Note: The diamond boat dimensions are approximately 3.5×2×1.2 cm. We typically mill away ∼3mm on each side and leave ∼4mm of aluminum flanking the diamond knife.

CRITICAL: Before machining, protect the diamond edge with paraffin wax to prevent debris damage. Failure to do so will likely result in aluminum shards damaging the knife edge. After milling, remove paraffin by peeling, then soak the knife overnight in soapy water (Dawn dish soap), rinse, and dry before use. Alternatively, clean the knife gently with ethanol and a Styrofoam cleaning stick provided with the knife upon purchase.

• 3.
  Mount the machined knife in the ultramicrotome (Leica UC7) at a 6° clearance angle.

  • a.
    Place the 3D-printed wafer attachment into position on the knife boat. If the milling was precise to the design of the wafer attachment, it should fit snuggly enough to hold in place.

Note: If not, you can temporarily secure it with 3M Transpore Medical Tape (Fisher Scientific, Cat. No. 50-283-1145), or modify the CAD diagram of the wafer attachment to better fit your milled knife.

• 4.
  Place a miniature bubble level (Fisher Scientific, Cat. No. 03-500-270) in the wafer holder recess and level the attachment both North–South and East–West by adjusting its position (Methods Video S1).

  • a.
    Once level, permanently seal the attachment with a hot glue gun or any other desirable, water impermeant, adhesive.
  • b.
    Allow full cure.

CRITICAL: Leveling is essential: any tilt causes floating sections to drift to one side of the wafer. Verify level at 6° knife clearance angle, as this is the standard sectioning angle.

• 5.
  Fill the completed knife boat assembly with MilliQ water and monitor for leaks over 24 h.

  • a.
    Seal any leaks with additional hot glue or adhesive and re-test.

Pause point: The assembled knife boat can be stored dry indefinitely once verified as watertight.

Figure 1.

Preparation of the wafer attachment and knife boat

(A) Wafer attachment is 3D printed using supplied CAD diagrams. Note the empty space at the top to fit the knife.

(B) A commercial diamond knife is shown (left) in both profile and overhead view. The boat is milled down (right) at a 6-degree clearance angle relative to the knife, avoiding the set screw.

Methods Video S1. Leveling the wafer attachment

Using the miniature bubble level, adjust the attachment first in the East-West direction (with the knife blade being the North direction), ensuring complete leveling. Then, adjust in the North-South direction, leaving the bubble slightly skewed towards the South, so the water is slightly away from the knife.

Preparation 2: Prepare plasma-cleaned silicon wafers with thermoplastic barriers

Timing: ∼60 min

In this preparation, 100 mm silicon wafers are made hydrophilic by O2 plasma treatment, then fitted with hydrophobic thermoplastic barriers that will confine sections over the wafer surface during sectioning.

• 6.
  Clean 100 mm silicon wafers (University Wafer, ID 1196) using an O2 plasma system (e.g., Plasma Etch PE-100LF: 100 W, 1 min, or equivalent instrument).

  • a.
    Store cleaned wafers in antistatic bags (Fisher Scientific, Cat. No. NC2388669) until used to prevent static buildup that renders them hydrophobic.

CRITICAL: Place the O2-cleaned wafer into water within 30–60 min of plasma treatment if not storing in antistatic bag. Static charge can build up on the wafer surface and reverse hydrophilicity. Glow-discharge systems provide equivalent results if a plasma etcher is unavailable. Before using the wafer to collect sections, test its hydrophilicity by depositing a drop of water on it. If the water spreads out upon impact, it’s still hydrophilic and ready to use. If the drop holds its surface tension (e.g., remains as a droplet on the silicon), then you will need to plasma clean the silicon.

Note: We have not extensively tested how long an O2-cleaned wafer will maintain its hydrophilicity when stored in an antistatic bag, as we routinely use them immediately. Using the water drop test described above is the best way to ensure the wafer is ready to use. However, we have found that storing the O2-cleaned wafer in water, in the knife boat, will maintain hydrophilicity for over a week.

• 7.
  Place the cleaned wafer into the knife boat assembly with the flat edge of the wafer facing the knife. This ensures the wafer fits as close to the knife as possible.

  • a.
    If using a hot-glue gun to build a barrier, we use standard ethylene-vinyl acetate (EVA) hot-melt glue sticks.
  • b.
    First make a shallow ring around the edge of the silicon wafer, let it set for ∼30 s (Methods Video S2).
  • c.
    Then go back over it with another layer, gradually building a wall that is taller than the wafer boat. This will ensure the barrier is taller than the layer of water you will use for sectioning.
  • d.
    Leave the flat edge of the wafer (the side that will face the knife) open to allow section entry.

Note: For a reusable alternative, affix a 3D-printed ASA barrier ring to the wafer holder. Place a small dot of hot glue on the two open ends of the barrier ring and affix to the wafer holder flanking the knife (Figure 2). The ASA ring can be lifted and reused between collection runs. A hybrid approach - ASA ring sealed along its inner perimeter with a thin bead of hot glue EVA - combines the precision of the printed ring with the reliable sealing of the hot-glue.

Note: EVA hot glue adheres well to dry silicon and remains attached when submerged. It detaches cleanly with forceps after wafer removal and poses minimal risk of disturbing sections during removal.

Figure 2.

Attachment points for ASA 3D-printed barrier

Barrier endpoints (green outline) are attached to the modified wafer boat nearest the remaining aluminum of the knife (blue) using hot glue (red circles).

Methods Video S2. Making the barrier

Once the wafer attachment is leveled, use a hot glue fun to seal the edge, ensuring the glue does not trail into the middle of the boat. The best method for building the barrier is to pump out the glue and move the gun towards the outside of the attachment.

Preparation 3: Mount and adjust ionizers

Timing: ∼10–15 min (initial setup; negligible thereafter)

Static ionizers generate the laminar EHD flow that clears sections away from the knife edge. Initial placement requires brief empirical tuning but need not be adjusted between runs once optimized.

• 8.This original design utilized 3D-printed articulated arm mounts to the ultramicrotome turret. However, commercially available articulating arm attached to an alligator clamp (see key resources table) works well and allows for precise positioning.
• 9.
  Position the two ionizers approximately 2 to 3 cm above the water surface, on either side of the knife, and directed toward the knife edge.

  • a.
    Ionizer positioning and water flow can be observed by placing a few small Styrofoam beads (manually scraped off a standard Conical tube rack) on the water and then removed after the desired flow is achieved.
  • b.
    A single ionizer positioned near the ultramicrotome chuck and pointed slightly away from the ultramicrotome readily moves sections away from the knife even after >1000 sections have been cut (Figure 3 and Methods Video S3). One ionizer is the minimum required.

Note: Once an effective position is found, it’s recommended to mark the ionizer position or make a note for future reference.

Note: The choice of two ionizers near the knife edge or one ionizer near the knife chuck is purely due to individual lab constraints and desires. The choice may be influenced by the specific ultramicrotome used (see materials and equipment). We have found no difference in performance, neither in dual vs single setup, nor in relation to block face size.

Figure 3.

Overall DRIFT-EM setup

Static ionizer is attached using a silicone flexible arm and alligator clip (left object with large circular black base). Ionizer tip (white/silver end coming off alligator clip) is positioned left of the ultramicrotome sample holder and facing slightly inward towards the knife boat. Wafer-modified boat is held in a standard Leica UC7 ultramicrotome knife holder. Wafer boat depicts a combined ASA 3D-printed barrier plus a thin layer of hot glue around the inner perimeter of the barrier to ensure a 100% seal around the silicon.

Methods Video S3. DRIFT-EM collection

Movie showing DRIFT-EM with ionizer placement as outlined in Figure 2 and after ∼1500 sections were collected. Ionizer effectively “pushes” sections (small silver objects) away from the knife edge as they accumulate towards the back of the boat.

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Biological samples
Adult C57BL/6 mouse brain (chemically fixed, heavy-metal stained, epoxy-embedded tissue block)	In-house preparation	N/A
Chemicals, peptides, and recombinant proteins
Ethylene-vinyl acetate (EVA) hot-melt glue sticks	Standard hardware supplier (e.g., Elmers, Gorilla glue)	N/A; any brand
3M Transpore Medical Tape	Fisher Scientific	Cat. No. 50–283–1145
Paraffin wax	Standard laboratory supply	N/A; any brand
Isoflurane, USP	Covetrus	11695–6777–1
0.2M Sodium Cacodylate Buffer	Electron Microscopy Sciences (EMS)	11653
16% Paraformaldehyde	EMS	15710
25% Glutaraldehyde	EMS	16220
LX-112 Resin	EMS	14900
EMbed 812 Resin	Ladd Research	21310
PLA	Dynamism, Inc.	DYN-PLA1-BK750G
ASA	Dynamism, Inc.	PF01044
Dish soap	Dawn	08285
Software and algorithms
DRIFT-EM ROI detection, nearest-neighbor ordering, and image normalization pipeline (Python)	This paper; Medina and Gogola, et al.1	https://github.com/gwildenberg/DRIFT-EM/tree/v1.0.0; https://doi.org/10.5281/zenodo.21972788
Sparse graph reconstruction and seriation	Yang et al.8	https://github.com/fumingyang-felix/sparse-graph-reconstruction-and-seriation-for-large-scale-image-stacks; https://doi.org/10.48550/arXiv.2509.23084
Zeiss ATLAS software	Zeiss	https://www.zeiss.com/microscopy
ImageJ/FIJI with MagC plugin	Templier6	RRID: SCR_002285, https://github.com/templiert/MagC
TrakEM2 (FIJI plugin) or any other EM alignment software	Cardona et al.9	RRID: SCR_008954, https://imagej.net/plugins/trakem2
AlignTK	Bock et al.10	https://mmbios.org/aligntk-home https://github.com/gwildenberg/aligntk_scripts.git
CAD software	Tinkercad	https://www.tinkercad.com
Other
100 mm silicon wafers	University Wafer	ID 1196
Diatome Ultra-Maxi diamond knife	Diatome	https://www.diatomeknives.com/ultra-maxi
Trim90 diamond knife	Diatome	https://www.diatomeknives.com/trim-90
Leica UC7 ultramicrotome	Leica Microsystems	RRID: SCR_016694
Static ionizers (2)	Diatome	Static Line II Ionizer (Set)
Ionizer clamp	Snakeclamp	1. Round Base Weighted: SCP-RBW 2. 13″ Flexible Gooseneck Tube Arm: SCP-GN13SB 3. Micro-clamp: SCP-MICRO
Plasma etcher (O2 plasma)	Plasma Etch	PE-100LF
Zeiss Gemini 300 SEM with airlock	Zeiss	RRID: SCR_023607
Zeiss Axio Imager Vario (optical microscope)	Zeiss	RRID: SCR_027233
3D printer	Bambu Lab	e.g., X1C 3D printer
Vibratome	Leica	VT1200S
Hot-glue gun	Gorilla	100424
Transfer pipette	Fisher Scientific	13-711-9BM
Eyelash brush	Ted Pella	113
2-flute carbide endmill	Grainger	19LC82
Miniature bubble level	e.g., Amazon ASIN B0DY7WP2S1	Fisher Scientific Cat. No. 03–500–270
4-inch wafer tweezers	Ted Pella	5048-SV, or equivalent
Large-gauge syringe	Standard laboratory supply	N/A
60°C hotplate with cover	Standard laboratory supply	N/A
Antistatic bags	Fisher Scientific	Cat. No. NC2388669

Materials and equipment

The following notes supplement the key resources table with additional equipment settings and alternatives.

• •Ionizer voltage: Typical operating range is ±4–8 kV. Start low (∼±4 kV) and increase if sections do not clear from the knife edge within a few seconds of cutting.
• •Ionizer type: Any commercial static ionizer that produces a balanced stream of positive and negative ions is suitable. The key requirement is the ability to position the emitter head 2–3 cm above the water surface.
• •Barrier alternatives: EVA hot-melt glue (any brand; clear or colored) works well. ASA 3D-printed barriers provide greater geometric precision and are reusable. A hybrid approach (ASA ring sealed with EVA at the silicon interface) is recommended when the ASA ring does not sit perfectly flush on the wafer.
• •Plasma cleaning alternatives: Any O2 plasma or UV-ozone system that renders the silicon surface hydrophilic is suitable. Glow-discharge systems commonly found in EM laboratories or Chemistry clean rooms are effective. Aim to submerge the cleaned wafer in water within 30–60 min of treatment.
• •Ultramicrotome alternatives: Any ultramicrotome capable of accommodating a standard Diatome Ultra-Maxi knife boat should be compatible with the modified knife boat, assuming the standard lower shank mount has not been modified. Certain versions of the Powertome (RMC) with integrated HD video systems may limit clearance near the knife edge, necessitating the single off-center ionizer format described above (Preparation 3).
• •SEM alternatives: Any SEM capable of accommodating 100 mm wafers and equipped with backscatter detection can be used. The ATLAS-compatible .proj export from the ROI pipeline is specific to Zeiss ATLAS software; adaptation to other SEM acquisition software requires minor code modifications (see GitHub repository).
• •Tissue alternatives: The below describe preparation of adult C57BL/6 male postnatal day 60 mouse brains for this protocol. This was a deliberate choice, as it is the age, sex, strain, and species from which most of our lab’s data has historically been collected, and was useful to evaluate the reproducibility of the protocol in relation to our standard workflows. In principle, this protocol is compatible with any typical sample type that you have already validated in-house to uncover your features of interest on an electron microscope.

Fixative (pH 7.4)

Reagent	Final concentration	Amount (mL)
Paraformaldehyde	2%	3.125
Glutaraldehyde	2.5%	2.5
Cacodylate Buffer	0.1 M	12.5
MilliQ H2O	N/A	6.875
Total	N/A	25 mL

Prepare fresh on ice, and perfuse ice-cold. Use immediately and do not store long-term.

Step-by-step method details

Tissue preparation and block trimming

Timing: ∼3–5 days (standard EM sample preparation)

This section covers the preparation of epoxy-embedded tissue blocks from mouse brain. If tissue blocks are already available from prior sample preparation, proceed to the next major step.

• 1.Deeply anesthetize adult C57BL/6 mice (male, postnatal day 60) with isoflurane and confirm loss of reflexes.

Note: All procedures must be performed under IACUC approval. Consult your institutional animal care committee for species- and age-specific anesthesia dosing.

Note: Age, sex, and strain should not be considered essential variables to protocol success.

• 2.Transcardially perfuse with 0.1 M sodium cacodylate buffer (pH 7.4), followed by fixative. Extract the brain immediately after perfusion.

Note: Make both solutions fresh immediately before use, store on ice, and perfuse ice-cold. Do not store working solutions long-term.

CRITICAL: Aim to complete perfusion within 5–10 min of anesthesia. A slow flow rate of buffer and fixative (∼3 to 4 mL/min) is essential to preserve cellular membranes and ultrastructure.

• 3.Post-fix brains in the same fixative for 24 h at 4°C. Rinse briefly in 0.1 M sodium cacodylate buffer, pH 7.4.

Pause point: Fixed brains can be stored in 0.1 M sodium cacodylate buffer at 4°C for several days to weeks before further processing.

• 4.Vibratome-section fixed brains at desired thickness (e.g., 100–200 μm). Cut out a desired region of interest from the correct coronal section.
• 5.
  Stain selected sections with heavy metals.

  • a.
    Dehydrate and embed in epoxy resin (LX-112 or Epon) using standard protocols.²
  • b.
    Trim embedded samples to desired area. ∼1 mm² works well for serial sectioning. Use a razor blade and/or a Diatome Trim90 knife for trimming.
  • c.
    Once the area is trimmed, use a Trim90 knife to produce a flat surface on the face of the tissue sample.

Note: Be sure the Trim90 tool is free of damage to avoid causing scratches or imperfections on the face of the tissue block.

Note: Full sample preparation (i.e., heavy-metal staining and resin embedding) is considered outside the scope of this protocol. Please refer to Hua et al.² for details.

Note: In principle, this protocol should be compatible with any typical sample source, heavy-metal staining preparation, or resin embedding media choice that you have already validated in-house to uncover your features of interest on an electron microscope.

Silicon wafer placement and water fill

Timing: ∼30 min

In this step, the prepared silicon wafer with thermoplastic barrier is placed into the knife boat, filled with water, and ionizers are positioned for the collection run.

• 6.Remove the plasma-cleaned silicon wafer from its antistatic bag and place the wafer into the 3D-printed wafer holder in the knife boat assembly ensuring the flat end of the wafer is positioned next to the knife.
• 7.
  If using a hot glue barrier, build barrier around the perimeter of the wafer making sure to leave a gap where the knife is.

  • a.
    If using a 3D-printed barrier that’s been previously attached to the boat, use a thin layer of hot glue to seal the barrier to the silicon wafer to ensure there are no small gaps where sections can slide through when removing the water after sectioning.
  • b.
    Fill the wafer boat with MilliQ water to the standard level used for ultramicrotomy (typically such that the wafer surface is just below the meniscus level).
  • c.
    Adjust the water level following your normal knife setup procedure.
• 8.
  Position the ionizer(s) on the articulated arm mounts at the pre-optimized angles (see Preparation 3).

  • a.
    Power on the ionizers and confirm gentle water movement away from the knife edge using dust particles or a small piece of styrofoam on the water surface.

CRITICAL: Verify that the ionizer flow clears away from the knife edge - not toward it - before beginning to section. Sections swept toward the knife will accumulate and cause pileups. If uncertain, refer to Supplementary Videos 6–7 from Medina and Gogola, et al.¹ for examples of centripetal vs. centrifugal flow.

Serial section collection

Timing: Variable; ∼1 h per 500 sections at 1–1.2 mm/s cut speed

This major step covers the actual collection of ultrathin sections onto the wafer. Sections are cut using standard ultramicrotomy technique; the ionizers handle clearing automatically.

• 9.Set ultramicrotome cutting parameters: section thickness 40 nm, cutting speed 1.0–1.2 mm/s.
• 10.Begin serial sectioning. Observe that each section floats off the knife edge and is cleared by the ionizer flow toward the back of the wafer within a few seconds of release.

Note: If sections pile up at the knife edge, increase ionizer voltage by 1–2 kV increments and/or adjust the ionizer angle. A single off-center ionizer generating circular flow is often sufficient.

• 11.Periodically monitor the water level every 10–20 min (frequency depends on room humidity and temperature). Add a few drops of ultrapure water to the back of the wafer holder to maintain a consistent water level.
• 12.Continue sectioning until the desired number of sections is reached (typically 2,000–3,000 sections per 100 mm wafer for 1 mm² blockface) or the wafer becomes full. Stop the microtome and turn off the ionizer.

Note: Continuous sessions of up to 3,000 sections have been validated without manual intervention for section clearing. Typically, sectioning is stopped when the density gets high enough where sections can no longer be cleared from the knife area by the ionizer. Adjusting ionizer voltage can compensate for temporary congestion near the cutting area.

Water removal and wafer drying

Timing: ∼45–60 min

Controlled water removal allows sections to settle and adhere to the silicon wafer surface without wrinkling or displacement.

• 13.After stopping the microtome, if any sections remain near the knife edge, gently move them away using a light puff of air from a plastic transfer pipette.
• 14.
  Prepare a small EVA plug: deposit a small amount of EVA hot glue onto a work surface.

  • a.
    Allow to cool, and cut to ∼4–5 mm width (matching the gap in the barrier near the knife).
  • b.
    Once cooled and firm, carefully place the plug in the barrier gap to seal the barrier ring (Figure 4).

CRITICAL: Sealing the barrier ring before water removal is essential to prevent sections from drifting off the wafer edge during the removal process. Allow the plug to stabilize for ∼1 min before proceeding.

Note: Sections will often rush away from the EVA plug when it contacts the water. This is normal and to be expected given the repulsive nature EVA has on ultrathin sections. If it’s desired to have the sections more evenly distributed across the wafer, turn the ionizer back on for a few sections to move the sections until a desired redistribution is met.

• 15.
  Using a large-gauge syringe or plastic transfer pipette, slowly withdraw water from the back of the wafer boat.

  Note: Sections might move depending on how quickly the water is removed (Methods Video S4).

  • a.
    Monitor the knife/EVA plug area to ensure the plug is keeping sections from falling off the wafer.
  • b.
    Remove water gradually to avoid disturbing sections.
  • c.
    Allow residual water to evaporate passively (∼20 min).

    Note: An alternative method for water removal is to modify the wafer CAD to have a small hole in the back of the wafer boat and connect tubing to the hole that is controlled by a peristaltic pump. Program the pump to pull water from the boat at a fixed rate.

• 16.
  Once sections are stationary and the wafer surface appears dry, use 4-inch wafer tweezers to lift the silicon wafer from the boat.

  • a.
    If using a hot glue barrier, pick up the wafer with the barrier still attached.
  • b.
    If using a 3D printed barrier, gently lift the barrier from the back side with one forceps while carefully removing the wafer with a second.
  • c.
    If properly attached to the front side, the barrier will stay attached to the boat and be ready for the next wafer if more sectioning is desired.

Note: When using the 3D printed barrier, there may be residual water around the perimeter of the wafer. Prop up the barrier by placing a small stick (e.g., eyelash brush) in between the barrier and the wafer boat to hold the barrier above the wafer and let the remaining water evaporate (Figure 5).

• 17.Place the wafer on a 30–60°C hotplate for 30–60 min, covered to prevent dust contamination. This completes drying and ensures sections adhere firmly to the wafer (Methods Video S5).

Pause point: Dried wafers can be stored in a clean, covered container at room temperature indefinitely before SEM imaging.

Figure 4.

Barrier plug to seal perimeter of silicon wafer

After sections are collected and still floating on water, small barrier plug made of hot glue (red outline) is placed between the two ends of the ASA barrier (green outlines) to create a complete seal around the silicon wafer. After the barrier plug is placed, water is slowly removed to allow sections (collection of silver rectangles at bottom of image) to passively attached to the silicon wafer.

Figure 5.

Wafer drying with an ASA barrier

After majority of water is removed, the ASA barrier is carefully lifted with forceps and propped up above the silicon wafer using an eyelash paintbrush. The ASA barrier remains attached to the wafer boat at the tips near the knife (see Figure 1). A-B show different angles of the setup.

Methods Video S4. Water withdrawal

An EVA hot glue plus is made to fit in the knife-edge gap to seal it once collection is finished. Water is gently removed from the back, non-knife, side of the wafer boat. A correctly leveled attachment (Preparation Step 4; Methods Video S1) should promote section movement away from the knife during water removal. Slow down water removal or turn on the ionizer if you notice sections gathering toward the plug. Monitor the plug to ensure sections do not fall off the wafer.

Methods Video S5. Drying the wafer

Place the wafer on the hot plate and allow the remaining water to dry. This could take anywhere from 10–60 minutes. If more time is needed, cover the wafer (e.g., with an inverted petri dish) to protect from dust.

Automated ROI detection and mapping

Timing: ∼30–60 min (local workstation processing)

A custom Python pipeline processes an optical overview image of the wafer to automatically detect, order, and export regions of interest (ROIs) for automated SEM imaging.

• 18.
  Collect a high-resolution optical overview image (Figure 6A) of the entire wafer surface using a widefield microscope (e.g., Zeiss Axio Imager Vario with Zen software).

  • a.
    A 2.5×/0.06 or 5×/0.13 HD EC Eliplan-NEOFLUAR objective is sufficient.
  • b.
    Save as a TIFF file.

Note: The resulting TIFF file should have a pixel size of approximately 7 microns, and be approximately 400–800 MB in size, depending on the packing density of sections.

• 19.
  Run the automatic ROI detection script from the DRIFT-EM repository (https://github.com/gwildenberg/DRIFTEM/tree/v1.0.) on the optical overview image. The four-stage pipeline performs:

  Note: Preprocessing: contrast-limited adaptive histogram equalization (CLAHE, 8×8 tiles, clip limit = 3.0).

  • a.
    Edge detection: Sobel operators (5×5 kernels) applied to horizontal and vertical gradients, followed by magnitude calculation and low-threshold binarization (threshold = 20).
  • b.
    Contour extraction: binary threshold = 86; dilated edges (4×4 kernel, 3 iterations) subtracted from binary image to isolate sections; external contour detection.
  • c.
    ROI optimization: Section dimensions are a required user input rather than a fixed value to accommodate for different choices, specified either in pixels or in microns together with the image scale.

    • i.
      The ROI outline may additionally be specified as a rectangle, square, trapezoid, ellipse, traced polygon, or the per-section convex hull.
    • ii.
      For each candidate contour, the initial pose is taken from its minimum-area bounding rectangle, then refined by L-BFGS-B minimization of mean intensity within the template under bounded translation (default ±5 pixels) and rotation (default ±90°), with no spatial overlap permitted between accepted ROIs.
    • iii.
      For outlines lacking 180° symmetry the optimization is run from both orientations and the better fit retained. Because accepted ROIs may not overlap, specifying a template slightly smaller than the true section increases yield on densely packed wafers.
    • iv.
      Recorded ROIs are therefore optionally enlarged after placement, by a per-axis margin or a fixed percentage, so that residual error in registering the optical overview to the SEM stage does not clip the section.

      Note: The algorithm correctly identifies ROIs (Figure 6B) for ∼91% of sections. Results are exported as .magc coordinate files with 10× scaled x,y coordinates of section locations, for compatibility with the legacy MagC plugin for downstream processing.

• 20.Open the .magc output file in ImageJ/FIJI using the MagC plugin (https://github.com/templiert/MagC). Manually inspect all ROIs and add bounding boxes for any missed sections using the manual stamping tool (∼10% of sections on average).
• 21.
  Run the nearest-neighbor ordering script to computationally reorder ROIs to minimize stage travel distance between consecutive imaging targets.

  • a.
    Export the ordered ROI list as a Zeiss ATLAS-compatible .proj XML file.

Note: The ATLAS .proj export mirrors the section-naming nomenclature of ATLAS files. For use with other SEM acquisition software, minor script modifications are required - see the repository README for guidance.

• 22.Load the .proj file into Zeiss ATLAS software on the SEM workstation. Review and confirm the acquisition order before beginning automated imaging.

Note: You may have to work with your EM core facility to make the output of the ROI finder compatible with their SEM.

Figure 6.

Optical overview

The .magc output file shows an example of completed ROI detection and mapping. A high-resolution overview is first collected (A; Step 18), followed by ROI detection (B, blue bounding boxes; Step 19) and algorithmic section ordering (B, inset numbers; Step 21).

Automated SEM imaging and image processing

Timing: Variable; depending on ROI size, desired dwell time and resolution. On average, 1–5 days.

This step covers automated wafer loading, SEM data acquisition, and post-acquisition image processing to produce an ordered, aligned 3D volume.

• 23.Load the wafer into the Zeiss Gemini 300 SEM airlock and transfer to the imaging chamber. Allow vacuum to stabilize.
• 24.In Zeiss ATLAS software, open the .proj file. Set acquisition parameters and begin automated acquisition.

Note: All image processing occurs on a local workstation before wafer loading. This minimizes SEM occupancy time — a practical advantage for shared core-facility instruments.

• 25.After acquisition, apply image normalization using the open-source normalization scripts from the DRIFT-EM repository (https://github.com/gwildenberg/DRIFT-EM/tree/v1.0.0; Zenodo: https://doi.org/10.5281/zenodo.21972788).
• 26.Perform 2D-image stitching if imaged ROI requires tiling using TrakEM2 (FIJI plugin; Cardona et al.⁹) or equivalent software.
• 27.Perform seriation to reorder images by biological continuity using the sparse graph reconstruction algorithm⁸ (https://github.com/FumingYang-Felix/Sparse-Graph-Reconstruction-and-Seriation-for-Large-Scale-Image-Stacks).
• 28.
  Perform 3D alignment of the ordered image stack using TrakEM2 (FIJI plugin; Cardona et al.⁹) or equivalent alignment software.

  • a.
    We find AlignTK (MMBIOS) works well for aligning EM images at scale for users who do not desire a graphical interface. Source code for installing and running AlignTK can be found here: https://github.com/gwildenberg/aligntk_scripts.git.

Expected outcomes

A successful DRIFT-EM collection run with a 1 mm² blockface at 40 nm section thickness should yield approximately 2,000–3,000 sections per 100 mm silicon wafer. Sections should be evenly distributed across the central ∼90 mm diameter collection zone (corresponding to approximately 6,362 mm²), with a mean distance of ∼5.3 ± 0.14 mm from the wafer edge. Individual sections should lie flat, unwrinkled, and well-separated without stacking on top of each other.

The automated ROI pipeline should detect bounding boxes for approximately 91% of sections from a single optical overview image. After manual correction (∼10 min), the complete ordered ROI set is exported as a Zeiss ATLAS-compatible.proj file, enabling automated SEM imaging without further user intervention until acquisition is complete.

At standard mapping resolution (55 nm/px lateral, 40 nm axial), the protocol produces 3D volumes spanning up to ∼200 × 200 × 74 μm (∼1,840 sections) with quality sufficient for automated neural circuit reconstruction. At higher resolution (6 nm/px), smaller volumes (e.g., 48 × 54 × 13 μm) resolve individual synapses, with pre-synaptic boutons, vesicle clusters, and post-synaptic densities clearly visible. Section morphology is equivalent to standard tape-based vEM; the ionizer introduces no observable artifacts. See Figures 4 and 5 of Medina and Gogola, et al.¹ for representative outcomes.

Quantification and statistical analysis

No statistical procedures are required for routine use of this protocol. Section collection yield (number of sections per wafer and collection zone geometry) can be quantified from the optical overview image and the ROI detection output. In our validation experiments (n = 3 independent collection runs), the mean distance between sections and the wafer edge was 5.3 ± 0.14 mm (mean ± SEM), corresponding to a collection zone diameter of approximately 90 mm. Section packing density can be compared across conditions by measuring the number of sections detected per unit wafer area from the .magc output file.

Limitations

Several practical considerations should be noted for routine use. First, optimal ionizer placement requires brief empirical tuning. While the clearing effect operates over a broad range of positions and voltages, initial users should expect to spend time finding an effective configuration for their specific microtome and ionizer model. Second, constructing the thermoplastic EVA barrier requires a hot-glue gun and moderate manual dexterity; minor inconsistencies in barrier height or shape can alter water dynamics and section distribution. The 3D-printed ASA barrier alternative reduces this variability. In principle, other plastic barriers could be implemented (e.g., using PDMS) so long as they are verified that that do not stick to sections. Third, knife modification requires access to a machine shop with milling capability; however, such facilities are standard at most university campuses. Fourth, the automated ROI pipeline currently exports Zeiss ATLAS-compatible .proj files only; adaptation for other SEM acquisition platforms (e.g., FEI/Thermo Fisher, JEOL) requires minor code modification, though this remains untested. Fifth, the maximum validated collection run is ∼3,000 sections on a single 100 mm wafer; wafer exchange to extend collection across multiple wafers has been tested in DRIFT-EM. The principal challenge to continuing to section on a second wafer is restarting the sectioning where the first sections could be partial sections due to deformations in the tissue sample, but this is independent of the core DRIFT-EM technology and a challenge shared across all direct pickup methods.

Troubleshooting

Problem 1: Sections accumulate at the knife edge and do not clear (related to Step-by-step #9–12)

Sections pile up near the knife rather than being swept toward the back of the boat.

Potential solution

• •Increase ionizer voltage in 1–2 kV increments (up to ±8 kV). Monitor section movement after each increment.
• •Adjust the ionizer angle to direct flow more directly away from the knife edge. Small angular changes (5–10°) can markedly improve clearing.
• •Verify that the ionizer is positioned above the water surface (not submerged or blocked). Ensure no metal or conductive objects are between the ionizer and the water surface.
• •If using a single ionizer, try repositioning it off-center to promote circular flow rather than bidirectional flow.

Problem 2: Sections wrinkle or fold when water is removed (related to Step-by-step #15–17)

Sections develop compression folds or wrinkles during or after water removal.

Potential solution

• •Ensure the wafer was O2 plasma cleaned within 30–60 min prior to use or stored in an antistatic bag. A hydrophobic wafer surface promotes wrinkling as the water recedes.
• •Slow the rate of water withdrawal. Remove water more gradually from the syringe.
• •Ensure the EVA barrier plug (Step 14) fully seals the opening before water removal to prevent turbulent edge effects.

Problem 3: Sections escape from the wafer area during sectioning (related to Step-by-step #8–11)

Sections are observed drifting outside the wafer boundary or over the barrier.

Potential solution

• •Inspect the EVA or ASA barrier for gaps or areas where height is below the waterline. Reinforce with additional EVA.
• •For the ASA ring, check that it sits flush on the wafer surface. If gaps exist between the ring and silicon, apply a thin bead of EVA along the inner perimeter (hybrid approach).
• •Reduce ionizer voltage slightly: at maximum voltage, the flow may be strong enough to push sections over a low barrier.

Problem 4: Water leaks from the knife boat assembly (related to Preparation no. 1, Step 5)

Water slowly drains from the boat during sectioning, requiring frequent refilling.

Potential solution

• •Remove the knife boat from the microtome and fill completely with water. Observe for 1–2 h to localize leaks.
• •Apply additional hot glue to leaking seams. Re-test for watertightness before proceeding.
• •Minor, slow water loss (∼1 mL/h) via evaporation is acceptable and can be compensated by periodic water addition (Step-by-step #12). Faster leakage must be sealed before collection.

Problem 5: ROI detection algorithm misses large number of sections (related to Step-by-step #19)

Automated ROI detection identifies fewer than ∼85% of sections in the optical overview image.

Potential solution

• •Adjust the CLAHE clip limit and binary threshold parameters in the initialization file (see repository README). Sections with low contrast against the silicon background may require parameter tuning.
• •Improve optical overview image quality: ensure even illumination of the wafer surface and collect at sufficient magnification to resolve individual sections.
• •For persistent detection problems, all sections can be manually stamped using the MagC ImageJ plugin (Step 20) as a complete fallback.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Gregg Wildenberg (gwildenberg@uchicago.edu).

Technical contact

Technical questions on executing this protocol should be directed to and will be answered by the technical contact, Gregg Wildenberg (gwildenberg@uchicago.edu).

Materials availability

STL files for all 3D-printed components are available at Zenodo: https://doi.org/10.5281/zenodo.19380076 and at https://github.com/gwildenberg/DRIFT-EM-CAD-files/tree/main.

Data and code availability

• •All datasets generated in this study have been deposited on Figshare and are publicly available at Figshare: https://doi.org/10.6084/m9.figshare.31836088.
• •All original code has been deposited on GitHub and is publicly available at Zenodo: https://doi.org/10.5281/zenodo.21972788 (algorithms) and Zenodo: https://doi.org/10.5281/zenodo.19380076 (CAD files).
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This project was supported by the NIH (1U01NS136401-01 and U01NS132158) and NSF Neuro Nex (no. 2014862).

Author contributions

Conceptualization, G.W.; methodology, J.V.G. and G.W.; visualization, J.V.G., A.L.J., and G.W.; writing – original draft and review and editing, J.V.G. and G.W.; funding acquisition, G.W.

Declaration of interests

The authors declare no competing interests.