Generating uniform tissue microfragments is important in many applications, including disease diagnostics, drug screening, spatial-omics, and fundamental wound healing and tissue regeneration studies. Common mechanical dissection methods, such as manual mincing, are imprecise and result in fragments with a broad range in size. This work aims to develop a microscale dicing device, referred to as the “μDicer,” consisting of a hollow array of blades spaced hundreds of micrometers apart. A tissue pushed through this array is diced into many microfragments simultaneously. The focus of this paper is on the fabrication process of the μDicer using a combination of isotropic and anisotropic etching in silicon. A single silicon oxide etch mask is used in a dry silicon etcher for both a tapered etch to form the microblades, and an anisotropic etch to form the through-holes in the hollow blade array. The use of a single mask reduces the mask fabrication time by more than twofold compared with two-mask approaches often used to generate similar etch features. The etch parameters and the design of the etch mask control the blade angles and the edge profiles of the blades. Specifically, the incorporation of “notches” in the two-dimensional mask design generates three-dimensional microserrated features on the blade edges. A custom, open-source etching model is also developed to facilitate the prediction of the etch profiles. Finally, a proof-of-concept application of the μDicer to dissect soft materials and tissues is demonstrated.

The ability to generate small and uniform fragments from tissues is important in many applications, including disease diagnostics (e.g., histology), drug screening (e.g., patient-derived organoids and personalized cancer medicine1,2), spatial-omics,3 wound repair, and regeneration studies.4,5 For some applications, mechanical dissection, instead of enzymatic digestion, is critical in preserving the tissue microenvironment and the in vivo association between different cell types, as it is less disruptive to the original tissue architecture.6 

Manual mincing of tissues with scissors or scalpels is the most common mechanical dissection method for preparing tissue fragments.7–11 However, the resulting fragments have a broad size range (∼10s–100s μm), which can lead to undesirable variability in subsequent analyses. In addition, the manual method is slow, labor-intensive and introduces contamination risk. A few studies have used tissue grinders or homogenizers that pass the tissue through a mesh filter and/or shear the tissue with a rotating blade.9,12 The resulting fragments still have a broad size distribution. The technique used to generate tissue microarrays has increased size uniformity, whereby a hollow needle is used to remove cylindrical plugs of tissue typically embedded in paraffin.13,14 The tissue plugs are then inserted into a recipient paraffin block in an array pattern. This block can be sectioned using a microtome for further analyses.15–17 However, this method is a serial process and cannot produce plugs ≲600 μm. In a recent study, a mechanical tissue chopper was used to dissect live biopsy specimens of mouse tumors into “cuboids” (300–600 μm in size) through a series of semi-automated chopping steps.2 While this method has higher throughput and fragment uniformity than some previous approaches, a key limitation is insufficient cutting. Incompletely cut cuboids require manual inspection and filtering. Laser capture microdissection (LCM) is by far the most precise tool for extracting cell clusters from a tissue slice.18–20 However, LCM requires the samples to be prepared in thin slices (typically 5–20 μm).18,21,22 Moreover, because LCM is a serial process, the isolation of many fragments is slow. Our group recently demonstrated a microfluidic “guillotine” to automate the bisection of organoids in a continuous flow manner.4,5 Although this method has a high cutting throughput, the current prototype can only cut in one plane. It would also be difficult to process larger samples (>1 mm) due to the small cross section of the microchannel.

The limitations of prior methods demonstrate a need for a better way to cut tissues into microfragments of controllable and uniform size. We aim to address this need by developing a microscale dicing device, referred to as the μDicer. Since the sectioning of tissues into thin uniform slices along the depth of the tissue (in the z direction) can be achieved by tools, such as the microtome, we focus on the dissection of tissue slices into uniform sub-millimeter fragments in the horizontal (x–y) plane. The μDicer consists of a hollow array of blades that are spaced hundreds of micrometers apart [Figs. 1(a) and 1(b)]. When a tissue is pushed through this array, it is diced into many microfragments simultaneously. In this paper, we focus on describing a microfabrication method to generate the μDicer in silicon using a combination of isotropic and anisotropic etching in a dry plasma etcher. Detailed characterization and application of the μDicer is the subject of a separate study.

FIG. 1.

(a) Diagram comparing our μDicer to manual mincing of tissue. (b) Photograph of the μDicer with a No. 10 surgeon's scalpel blade for reference. (c) The process flow for etching the blades and through-holes.

FIG. 1.

(a) Diagram comparing our μDicer to manual mincing of tissue. (b) Photograph of the μDicer with a No. 10 surgeon's scalpel blade for reference. (c) The process flow for etching the blades and through-holes.

Close modal

Needles and blades have been fabricated in silicon by isotropic etches or anisotropic self-limiting etches, but it is difficult to achieve small tip or blade angles.23 Previously, sharp, high aspect-ratio needles have been fabricated in silicon using tapered etching by cycling between an anisotropic Bosch process and a timed isotropic etch.24,25 Multiple etch masks were used to create the needles and the through-holes.24 The use of multiple lithographic masks and etching steps, often required for complex topographical etching, increases the process time and cost.24,25 While single masks have been used to fabricate 3D features,26 microprobe electrodes,27 and needles,28 no method has reported the use of a single mask for both a tapered etch followed by a through-hole etch to form blades. In addition, most of the previous methods used blades with straight edges. For efficient tissue cutting, the profile of the blade edge can be important. For example, microserrations on the blade have been found to reduce the force for cutting soft tissues.29,30 No work has reported the fabrication of blade with microserrations in silicon with a single etch mask.

In this paper, we demonstrate the fabrication of a hollow array of sharp blades with microserrations in silicon using a single lithographic and etch mask for two etch processes (i.e., tapered etch and through-hole etch). The use of a single mask reduces the mask fabrication time since the wafer does not require stripping and re-patterning. The etch parameters control the blade angles, whereas the notches on the etch mask control the microserrations along the blade. A custom etching model is developed to predict the etch profiles. Finally, we show a proof-of-concept application of the μDicer to dissect soft materials and tissues.

Figure 1(c) shows the fabrication of our μDicer. All etch parameters and details are listed in Fig. 2, Table S1, and Note S1 in the supplementary material. To form an array of microblades (step 1), we used a single etch mask composed of silicon dioxide and performed tapered etches by cycling between isotropic etches (for t1-iso seconds) and deep reactive-ion etches (DRIE) using a Bosch process (for n1 cycles). This process was repeated N rounds in an inductively coupled plasma deep silicon etcher (PlasmaTherm, Versaline). The values of n1 and t1-iso were chosen to control the angle of the overall etch and, thus, the half-angle of the blade [θ, Fig. 1(c)]. In general, θ is expected to decrease with the increasing ratio of anisotropic etch to isotropic etch. Experimentally, it has been shown that increasing n1 relative to t1-iso decreased θ, everything else held constant.25 Since the detailed examination was described previously,25 we did not repeat the characterization here. For the etch conditions tested, θ varied from approximately 8° to 34° [see Fig. 2(b)]. Here, θ did not decrease monotonically with the increasing ratio of n1 to t1-iso since we did not fix other variables, including the geometry of the mask, which changed the local etch rates in both the isotropic and anisotropic etches. The number of rounds (N) controlled the point at which the etch fronts undercutting the oxide mask met to form the blades. We chose N = 4 so that the etch stopped immediately before the etch fronts met. N >4 resulted in over-etching and instability of the mask. In step 2, we used the Bosch process (n2 cycles) to etch the through-holes most of the way through the wafer (∼350 μm). The depth of the etch was measured with digital microscopy at intermittent points. Additional Bosch cycles (in batches of 50) were performed until the remaining wafer thickness was ∼50 μm. In step 3, we sharpened the blades with an isotropic etch (t3-iso). The value of t3-iso was determined by monitoring the progression of the etch fronts every 100 s until the corner tips formed. After completing the etches, we removed the top oxide mask and ground the backside of the wafer until the through-holes were fully exposed. We chose to perform backside grinding instead of etching the wafer all the way because the latter resulted in unavoidable narrowing of the bottom of the through-hole, which would hinder the passing of the diced tissue.

FIG. 2.

(a) Diagram defining the blade geometries. (b) Table showing the corresponding mean measured values and device identifiers (IDs). n = 6 measurements were made on different blades within a single device with a given ID. Table S2 reports the standard deviations for each measurement.

FIG. 2.

(a) Diagram defining the blade geometries. (b) Table showing the corresponding mean measured values and device identifiers (IDs). n = 6 measurements were made on different blades within a single device with a given ID. Table S2 reports the standard deviations for each measurement.

Close modal

Our single-mask fabrication approach has two advantages. First, it reduced the mask fabrication time by >2 times compared with a two-mask approach. In our facilities, the mask fabrication process for a single wafer took ∼3 h. For a two-mask approach, this process would, thus, take at least 6 h, not accounting for additional time spent coating protective layers onto previously etched features. Second, the single-mask approach eliminated the need for mask alignment. Nevertheless, we recognize the trade-off in using a single etch mask, namely the limited control of the shape of the through-holes because the shape of the blades and through-holes were both determined by the same mask. Finally, we note that although it was possible to generate blades using isotropic etches only (in step 1), the blades were not sufficiently sharp (with θ ∼50°, Fig. S1 in the supplementary material) so we did not pursue this method further.

Figure 3 shows the effect of the shape of the etch mask on the formation of microserrations on the blade edges. Although the serrations were 3D features, a 2D etch mask design with notches was sufficient to create these serrations. Figure 3(a) shows our mask designs (only one cell shown) for μDicers with 0, 1, and 2 serrations, respectively (see Fig. S2 in the supplementary material for mask dimensions). The notches on the etch mask increased the distance between the two undercutting etch fronts from the neighboring cells and controlled the point at which the two etch fronts intersected [Fig. 3(b)]. As the etch proceeded, the two etch fronts intersected to result in a diamond-shaped plateau and eventually a tip at tf at the end of step 3 forming a serration on the blade edge. Here, tf corresponds to the combined isotropic etching time. Figure 3(c) shows SEM images of μDicers and the corresponding etch masks. As can be seen, the number of notches corresponded to the number of serrations. Figure 2 shows the measured number of serrations, serration heights, and blade angles for different etch mask designs and etch parameters used (also see Note S2 in the supplementary material). Figure 2 includes measurements of angles α and β, the angles of the corner, and serration tips that are coplanar with the blade edge (section b–b), respectively. They were controlled by the tapered etch (step 1), the mask design, and the isotropic etch (step 3) when the etch fronts from neighboring cells intersected under the oxide mask [also see profile at tf in Fig. 3(b)]. They are, thus, different from θ, which is orthogonal to the plane of the blade edge, in that θ depended primarily on the tapered etch. β scaled roughly with θ, but the trend in α was difficult to predict due to the non-trivial interactions of the etch fronts. Supplementary material Note S3 includes additional guidelines of the mask design.

FIG. 3.

(a) Diagram showing different mask designs and the corresponding serration geometries. Only a single cell of the entire array is shown. (b) A diagram showing the time evolution of the etch process showing how a serration is formed during steps 1 and 3 of the process flow. (Step 2 was excluded as it does not affect blade formation.) For simplicity, only two cells are shown. The asterisk indicates the expected profile at the end of step 1. Angles α, β, and θ were defined in Fig. 2(b). (c) SEM images of μDicers with 0, 1, and 2 serrations, respectively, with the corresponding device IDs and mask designs.

FIG. 3.

(a) Diagram showing different mask designs and the corresponding serration geometries. Only a single cell of the entire array is shown. (b) A diagram showing the time evolution of the etch process showing how a serration is formed during steps 1 and 3 of the process flow. (Step 2 was excluded as it does not affect blade formation.) For simplicity, only two cells are shown. The asterisk indicates the expected profile at the end of step 1. Angles α, β, and θ were defined in Fig. 2(b). (c) SEM images of μDicers with 0, 1, and 2 serrations, respectively, with the corresponding device IDs and mask designs.

Close modal

While the notches in the etch mask were able to generate 3D serrations, as expected, the resulting etch profiles were not always easy to predict. As such, we developed a custom model to simulate steps 1 and 3 of the etching process. The goal of this model was to provide a computationally inexpensive, qualitative prediction of the shape of the etched blade geometry, rather than to match the exact etch profile quantitatively. This simulation can, thus, inform design decisions before fabrication. Details of the model are described in Note S4 in supplementary material, and the code is available on GitHub.31 Briefly, the simulation implements a sparse field level set approach to model the etch fronts using an algorithm from Montoliu et al.32,33 The inputs to the model included the mask design file, the recipe steps, the known horizontal and vertical etch rates, and the desired resolution. The outputs of the model were isometric images of the etch profile produced every two time steps, and an interactive visualization (in .vtk format) of the final 3D model. By allowing open access to this code, we aim to make our model accessible for use and adaptation. The current version evolves the etch front using known etch rates, but the framework is in place to adapt the model to implement more complex etching physics (e.g., ray tracing, ion-surface chemistry, and other surface kinetics).34,35 The etch rates for our model were determined experimentally by performing an isotropic etch test on a patterned wafer and measuring the vertical and horizontal etch distances for a given etch time. Our model is adaptable for different combinations of isotropic and Bosch etches and can be applied to other plasma etchers with known etch rates. As shown in Figs. 4(b)–4(d), the model generated etch profiles that matched the experimental etch profiles [Fig. 3(c)] qualitatively, reproducing the correct number of serrations.

FIG. 4.

(a) Images depicting the progression of the etching process in steps 1 and 3 of the process flow simulated by our model for a mask with a single notch that generates a μDicer with one serration. Only four cells are modeled. Movie S1 shows an animation of this process. (b) Simulation results (for steps 1 and 3 of the process flow) for various mask designs included in the inset at the bottom left corner.

FIG. 4.

(a) Images depicting the progression of the etching process in steps 1 and 3 of the process flow simulated by our model for a mask with a single notch that generates a μDicer with one serration. Only four cells are modeled. Movie S1 shows an animation of this process. (b) Simulation results (for steps 1 and 3 of the process flow) for various mask designs included in the inset at the bottom left corner.

Close modal

As a proof of concept of its utility, we used the μDicer to cut agar (5% w/v), porcine articular cartilage, and porcine liver with approximate stiffnesses of 22.5 kPa,36 2.6 MPa,37 and 1.1 kPa,38 respectively. All samples (∼2–3 mm in diameter and 0.2–1 mm in thickness) were pressed and extruded through the μDicer (total blade array area ∼3 × 3 mm2) with a rubber tipped plunger (see the supplementary material, Fig. S3 for details). Figure 5 shows the uniformity of the fragments cut using the μDicer vs manual mincing. The median fragment size generated by different μDicers ranged from 187 to 206 μm, within 6.6% from the blade spacing of the respective μDicers used. In comparison, it was difficult to control the median fragment size using manual mincing. The maximum interquartile range (IQR) of fragments generated using μDicers was 43 μm (or 24% of the median) for the liver sample, whereas manual mincing had an IQR up to 127 μm (or 92.4% of the median) for the cartilage sample.

FIG. 5.

Images of (a) agar, (b) porcine cartilage, and (c) porcine liver cut by manual mincing (left column) and by μDicers (right column). The device IDs were included at the top left corner. (d) Boxplots of fragment widths generated by manual mincing vs μDicers with different number of serrations (m) and device IDs as labeled. The fragment width was determined by measuring the narrowest dimension passing through the centroid of the fragment in ImageJ. The box represents the interquartile range (IQR). The upper whisker is the largest dataset value smaller than 1.5 × IQR above the third quartile, and the lower whisker is the smallest dataset value larger than 1.5 × IQR below the first quartile. Black dots represent outliers. The horizontal black line indicates the median. The horizontal red line indicates the blade spacing (w) for the device used to cut that sample. Sample size (n) are included at the bottom of the plots. Fragments 20 μm were excluded here (see the text).

FIG. 5.

Images of (a) agar, (b) porcine cartilage, and (c) porcine liver cut by manual mincing (left column) and by μDicers (right column). The device IDs were included at the top left corner. (d) Boxplots of fragment widths generated by manual mincing vs μDicers with different number of serrations (m) and device IDs as labeled. The fragment width was determined by measuring the narrowest dimension passing through the centroid of the fragment in ImageJ. The box represents the interquartile range (IQR). The upper whisker is the largest dataset value smaller than 1.5 × IQR above the third quartile, and the lower whisker is the smallest dataset value larger than 1.5 × IQR below the first quartile. Black dots represent outliers. The horizontal black line indicates the median. The horizontal red line indicates the blade spacing (w) for the device used to cut that sample. Sample size (n) are included at the bottom of the plots. Fragments 20 μm were excluded here (see the text).

Close modal

The primary objective of these experiments was to compare fragment uniformity generated by the μDicers vs manual mincing. Since the uniformity did not vary significantly when dicing agar with μDicers having different blade geometries, we did not pursue to test cartilage and liver with identical devices. However, we expect that decreased blade angles (α, β, θ), increased number of serrations, and increased serration height relative to peak height would reduce the cutting force needed. Validating these effects requires measuring the force–displacement curves of these blades cutting into tissue. The relatively large ratio of peak height (p) to serration height (s) in our devices did not present challenges to the cutting process even though some tissues were thin (∼0.2 mm). It was likely due to the support from the soft plunger tip used to push the tissue through the μDicer. We note that our manual sample handling led to undesirable tissue fragmentation resulting in small fragments both from samples cut manually or by our μDicers. In our analysis, fragments less than ∼20 μm (1/10th of the blade spacing) were excluded since they were unrelated to the cutting by the μDicer and could be filtered out in practical use. Also, with the crude extrusion method here, some tissue remained trapped in the μDicer. The improvement in the method to load and extrude samples is a part of an on-going study to increase tissue recovery.

In summary, we have described the fabrication of μDicers in silicon using a single etch mask. The incorporation of notches in the 2D mask design generated complex 3D microserrations on the blades. The etching model facilitated the prediction of the etch profile qualitatively. The agar and tissue fragments cut by the μDicers were uniform and matched the blade spacing of the μDicers. While we have focused on a blade spacing of ∼200 μm in this work, our fabrication process should be able to generate blade spacing as small as ∼15 μm (from a 500 μm-thick silicon wafer), as limited by the plasma etching, which becomes challenging to create features with aspect ratios exceeding 30:1.39 The serrations made no significant difference in the fragment uniformity (tested in agar only), but are expected to change the cutting force. Ongoing work includes integrating the μDicers with automated sample loading and extrusion and the detailed characterization of the cutting process and its dependence on the blade geometries.

See the supplementary material for more details on etch parameters and the mask designs, the fabrication process, the etch simulation, alternative etching techniques we attempted, and calculations for determining the blade angles from SEM images. Additionally, a supplementary material movie animating the etch process is included.

This work was supported by funding from the Stanford Bio-X Interdisciplinary Initiatives Seed Grants Program (IIP) (No. R10-14); the Center for Cellular Construction, which is a Science and Technology Center funded by the National Science Foundation (NSF Award No. DBI-1548297); and class fabrication funds from ENGR 241 at Stanford. We thank Professor Roger Howe, Professor Jonathan Fan, Dr. Antonio Ricco, Dr. Mark Zdeblick, and Dr. Usha Raghuram for their guidance and helpful discussions. All devices were fabricated in the Stanford Nano Shared Facilities (SNSF), which is supported by the National Science Foundation under Award No. ECCS-2026822.

The data that support the findings of this study are available from the corresponding author upon reasonable request. The code for the etch model developed in this study is openly available on GitHub, Ref. 29.

1.
L. F.
Horowitz
,
A. D.
Rodriguez
,
Z.
Dereli-Korkut
,
R.
Lin
,
K.
Castro
,
A. M.
Mikheev
,
R. J.
Monnat
,
A.
Folch
, and
R. C.
Rostomily
,
npj Precis. Oncol.
4
,
12
(
2020
).
2.
L. F.
Horowitz
,
A. D.
Rodriguez
,
A.
Au-Yeung
,
K. W.
Bishop
,
L. A.
Barner
,
G.
Mishra
,
A.
Raman
,
P.
Delgado
,
J. T. C.
Liu
,
T. S.
Gujral
,
M.
Mehrabi
,
M.
Yang
,
R. H.
Pierce
, and
A.
Folch
,
Lab Chip
21
,
122
(
2020
).
3.
A. D.
Rodriguez
,
L. F.
Horowitz
,
K.
Castro
,
H.
Kenerson
,
N.
Bhattacharjee
,
G.
Gandhe
,
A.
Raman
,
R. J.
Monnat
,
R.
Yeung
,
R. C.
Rostomily
, and
A.
Folch
,
Lab Chip
20
,
1658
(
2020
).
4.
L. R.
Blauch
,
Y.
Gai
,
J. W.
Khor
,
P.
Sood
,
W. F.
Marshall
, and
S. K. Y.
Tang
,
Proc. Natl. Acad. Sci. U. S. A.
114
,
7283
(
2017
).
5.
S.
Toda
,
L. R.
Blauch
,
S. K. Y.
Tang
,
L.
Morsut
, and
W. A.
Lim
,
Science
361
,
156
(
2018
).
6.
J. T.
Neal
,
X.
Li
,
J.
Zhu
,
V.
Giangarra
,
C. L.
Grzeskowiak
,
J.
Ju
,
I. H.
Liu
,
S.-H.
Chiou
,
A. A.
Salahudeen
,
A. R.
Smith
,
B. C.
Deutsch
,
L.
Liao
,
A. J.
Zemek
,
F.
Zhao
,
K.
Karlsson
,
L. M.
Schultz
,
T. J.
Metzner
,
L. D.
Nadauld
,
Y.-Y.
Tseng
,
S.
Alkhairy
,
C.
Oh
,
P.
Keskula
,
D.
Mendoza-Villanueva
,
F. M.
De La Vega
,
P. L.
Kunz
,
J. C.
Liao
,
J. T.
Leppert
,
J. B.
Sunwoo
,
C.
Sabatti
,
J. S.
Boehm
,
W. C.
Hahn
,
G. X. Y.
Zheng
,
M. M.
Davis
, and
C. J.
Kuo
,
Cell
175
,
1972
(
2018
).
7.
X.
Li
,
L.
Nadauld
,
A.
Ootani
,
D. C.
Corney
,
R. K.
Pai
,
O.
Gevaert
,
M. A.
Cantrell
,
P. G.
Rack
,
J. T.
Neal
,
C. W.-M.
Chan
,
T.
Yeung
,
X.
Gong
,
J.
Yuan
,
J.
Wilhelmy
,
S.
Robine
,
L. D.
Attardi
,
S. K.
Plevritis
,
K. E.
Hung
,
C.-Z.
Chen
,
H. P.
Ji
, and
C. J.
Kuo
,
Nat. Med.
20
,
769
(
2014
).
8.
M. A.
Cantrell
and
C. J.
Kuo
,
Genome Med.
7
,
32
(
2015
).
9.
H.
Tamura
,
A.
Higa
,
H.
Hoshi
,
G.
Hiyama
,
N.
Takahashi
,
M.
Ryufuku
,
G.
Morisawa
,
Y.
Yanagisawa
,
E.
Ito
,
J.-I.
Imai
,
Y.
Dobashi
,
K.
Katahira
,
S.
Soeda
,
T.
Watanabe
,
K.
Fujimori
,
S.
Watanabe
, and
M.
Takagi
,
Oncol. Rep.
40
,
635
(
2018
).
10.
N. K.
Finnberg
,
P.
Gokare
,
A.
Lev
,
S. I.
Grivennikov
,
A. W.
MacFarlane
,
K. S.
Campbell
,
R. M.
Winters
,
K.
Kaputa
,
J. M.
Farma
,
A. E.-S.
Abbas
,
L.
Grasso
,
N. C.
Nicolaides
, and
W. S.
El-Deiry
,
Oncotarget
8
,
66747
(
2017
).
11.
R. L.
DiMarco
,
J.
Su
,
K. S.
Yan
,
R.
Dewi
,
C. J.
Kuo
, and
S. C.
Heilshorn
,
Integr. Biol.
6
,
127
(
2014
).
12.
V.
Baldan
,
R.
Griffiths
,
R. E.
Hawkins
, and
D. E.
Gilham
,
Br. J. Cancer
112
,
1510
(
2015
).
13.
G.
Sauter
,
R.
Simon
, and
K.
Hillan
,
Nat. Rev. Drug Discovery
2
,
962
(
2003
).
14.
R.
Simon
,
M.
Mirlacher
, and
G.
Sauter
,
BioTechniques
36
,
98
(
2004
).
15.
N. M. T.
Jawhar
,
Ann. Saudi Med.
29
,
123
(
2009
).
16.
L.
Fagerberg
,
B. M.
Hallström
,
P.
Oksvold
,
C.
Kampf
,
D.
Djureinovic
,
J.
Odeberg
,
M.
Habuka
,
S.
Tahmasebpoor
,
A.
Danielsson
,
K.
Edlund
,
A.
Asplund
,
E.
Sjöstedt
,
E.
Lundberg
,
C. A.-K.
Szigyarto
,
M.
Skogs
,
J. O.
Takanen
,
H.
Berling
,
H.
Tegel
,
J.
Mulder
,
P.
Nilsson
,
J. M.
Schwenk
,
C.
Lindskog
,
F.
Danielsson
,
A.
Mardinoglu
,
Å.
Sivertsson
,
K.
von Feilitzen
,
M.
Forsberg
,
M.
Zwahlen
,
I.
Olsson
,
S.
Navani
,
M.
Huss
,
J.
Nielsen
,
F.
Ponten
, and
M.
Uhlén
,
Mol. Cell. Proteomics
13
,
397
(
2014
).
17.
M.
Uhlén
,
L.
Fagerberg
,
B. M.
Hallström
,
C.
Lindskog
,
P.
Oksvold
,
A.
Mardinoglu
,
Å.
Sivertsson
,
C.
Kampf
,
E.
Sjöstedt
,
A.
Asplund
,
I.
Olsson
,
K.
Edlund
,
E.
Lundberg
,
S.
Navani
,
C. A.-K.
Szigyarto
,
J.
Odeberg
,
D.
Djureinovic
,
J. O.
Takanen
,
S.
Hober
,
T.
Alm
,
P.-H.
Edqvist
,
H.
Berling
,
H.
Tegel
,
J.
Mulder
,
J.
Rockberg
,
P.
Nilsson
,
J. M.
Schwenk
,
M.
Hamsten
,
K.
von Feilitzen
,
M.
Forsberg
,
L.
Persson
,
F.
Johansson
,
M.
Zwahlen
,
G.
von Heijne
,
J.
Nielsen
, and
F.
Pontén
,
Science
347
,
1260419
(
2015
).
18.
V.
Espina
,
J. D.
Wulfkuhle
,
V. S.
Calvert
,
A.
VanMeter
,
W.
Zhou
,
G.
Coukos
,
D. H.
Geho
,
E. F.
Petricoin
, and
L. A.
Liotta
,
Nat. Protoc.
1
,
586
(
2006
).
19.
R.
Longuespée
,
D.
Alberts
,
C.
Pottier
,
N.
Smargiasso
,
G.
Mazzucchelli
,
D.
Baiwir
,
M.
Kriegsmann
,
M.
Herfs
,
J.
Kriegsmann
,
P.
Delvenne
, and
E.
De Pauw
,
Methods
104
,
154
(
2016
).
20.
A.
Rabien
and
G.
Kristiansen
,
Methods Mol. Biol.
1381
,
39
(
2016
).
21.
S.
Datta
,
L.
Malhotra
,
R.
Dickerson
,
S.
Chaffee
,
C. K.
Sen
, and
S.
Roy
,
Histol. Histopathol.
30
,
1255
(
2015
).
22.
S.
Curran
,
J. A.
McKay
,
H. L.
McLeod
, and
G. I.
Murray
,
Mol. Pathol.
53
,
64
(
2000
).
23.
A.
Ganguli
,
A.
Ornob
,
N.
Spegazzini
,
Y.
Liu
,
G.
Damhorst
,
T.
Ghonge
,
B.
Thornton
,
C. J.
Konopka
,
W.
Dobrucki
,
S. E.
Clare
,
R.
Bhargava
,
A. M.
Smith
,
F.
Kosari
, and
R.
Bashir
,
Nat. Commun.
9
,
202
(
2018
).
24.
A.
Trautmann
,
F.
Heuck
,
R.
Denfeld
,
P.
Ruther
, and
O.
Paul
, in
19th IEEE International Conference on Micro Electro Mechanical Systems
(
IEEE
,
2006
), pp.
434
437
.
25.
N.
Roxhed
,
P.
Griss
, and
G.
Stemme
,
J. Micromech. Microeng.
17
,
1087
(
2007
).
26.
A.
Trautmann
,
P.
Ruther
, and
O.
Paul
, in IEEE
the Sixteenth Annual International Conference on Micro Electro Mechanical Systems
, MEMS-03 Kyoto (
IEEE
,
2003
), pp.
682
685
.
27.
Y.
Shin
,
Y.
Kim
,
S.
Lee
, and
J.
Park
,
Micro Nano Lett.
13
,
1245
(
2018
).
28.
S.
Negi
and
R.
Bhandari
,
Microsyst. Technol.
19
,
203
(
2013
).
29.
M.
Giovannini
,
H.
Ren
,
X.
Wang
, and
K.
Ehmann
,
J. Micro Nano-Manuf.
5
,
041004
(
2017
).
30.
J.
Lu
,
X.
Wang
,
Y.
Huang
,
C.
Zhou
,
B.
Xu
, and
Q.
Fu
,
Tribol. Int.
145
,
106162
(
2020
).
31.
See https://github.com/tanglab-stanford/PythonDryEtchModel for “
Python Tool for Modeling Silicon Etch Profile from the Dry Plasma Etching
” (
2021
).
32.
C.
Montoliu
,
N.
Ferrando
,
M. A.
Gosálvez
,
J.
Cerdá
, and
R. J.
Colom
,
J. Micromech. Microeng.
23
,
075017
(
2013
).
33.
C.
Montoliu
,
N.
Ferrando
,
M. A.
Gosálvez
,
J.
Cerdá
, and
R. J.
Colom
,
Comput. Phys. Commun.
184
,
2299
(
2013
).
34.
O.
Ertl
and
S.
Selberherr
,
ECS Trans.
23
,
61
(
2019
).
35.
H. H.
Hwang
,
T. R.
Govindan
, and
M.
Meyyappan
,
J. Electrochem. Soc.
146
,
1889
(
1999
).
36.
M. D.
Gaon
,
K. H. K.
Ho
, and
B. J. F.
Wong
,
Lasers Med. Sci.
18
,
148
(
2003
).
37.
M.
Stolz
,
R.
Raiteri
,
A. U.
Daniels
,
M. R.
VanLandingham
,
W.
Baschong
, and
U.
Aebi
,
Biophys. J.
86
,
3269
(
2004
).
38.
M.
Dosaev
,
I.
Goryacheva
,
Y.
Martynenko
,
A.
Morozov
,
F.
Antonov
,
F.-C.
Su
,
C.-H.
Yeh
, and
M.-S.
Ju
,
J. Med. Biol. Eng.
35
,
510
(
2015
).
39.
J.
Yeom
,
Y.
Wu
,
J. C.
Selby
, and
M. A.
Shannon
,
J. Vac. Sci. Technol. B
23
,
2319
(
2005
).

Supplementary Material