Fabrication of microfluidic devices by photolithography generally requires specialized training and access to a cleanroom. As an alternative, 3D printing enables cost-effective fabrication of microdevices with complex features that would be suitable for many biomedical applications. However, commonly used resins are cytotoxic and unsuitable for devices involving cells. Furthermore, 3D prints are generally refractory to elastomer polymerization such that they cannot be used as master molds for fabricating devices from polymers ( e.g. polydimethylsiloxane, or PDMS). Different post-print treatment strategies, such as heat curing, ultraviolet light exposure, and coating with silanes, have been explored to overcome these obstacles, but none have proven universally effective. Here, we show that deposition of a thin layer of parylene, a polymer commonly used for medical device applications, renders 3D prints biocompatible and allows them to be used as master molds for elastomeric device fabrication. When placed in culture dishes containing human neurons, regardless of resin type, uncoated 3D prints leached toxic material to yield complete cell death within 48 hours, whereas cells exhibited uniform viability and healthy morphology out to 21 days if the prints were coated with parylene. Diverse PDMS devices of different shapes and sizes were easily cast from parylene-coated 3D printed molds without any visible defects. As a proof-of-concept, we rapid prototyped and tested different types of PDMS devices, including triple chamber perfusion chips, droplet generators, and microwells. Overall, we suggest that the simplicity and reproducibility of this technique will make it attractive for fabricating traditional microdevices and rapid prototyping new designs. In particular, by minimizing user intervention on the fabrication and post-print treatment steps, our strategy could help make microfluidics more accessible to the biomedical research community.
more »
« less
Casting Octahedra for Reproducible Multi-Anvil Experiments by 3D-Printed Molds
Making consistent and precise octahedral pressure media is crucial for reproducible high-pressure experiments in the multi-anvil press. Here we report a new approach of casting octahedra using 3D-printed molds, and pressure calibrations for octahedra both with and without pre-existing gaskets (“fins”). The 3D-printed molds for casting octahedra from either Ceramacast 584-OF or 646 cement improve the reproducibility of the octahedra and allow for a pre-existing central hole (for the high-pressure cell assembly) in the final cast product. Pressure and temperature calibrations of the octahedra have been performed based on phase transitions in bismuth (Bi) and silica (SiO2), respectively, in order to determine the efficiency and reproducibility of pressure generation and thermal insulation for cast octahedra designed for use with 18/12, 14/8, and 10/5 multi-anvil assemblies. The pressure-generating efficiency of the 14/8 and 10/5 octahedra with pre-existing gaskets, cast from the 584-OF cement, is similar to that of the corresponding COMPRES (Consortium for Materials Properties Research in Earth Sciences) octahedra, and more efficient than pre-cast octahedra made from the same material but lacking pre-existing gaskets. The efficiency of pre-gasketed 18/12 octahedra made of the 646 cement is markedly lower than those of the 584 cement. However, the 18/12 large-volume octahedra, cast (with fins) from the ZrO2-based 646 cement, also provides efficient thermal insulation. Casting octahedral solid pressure media for multi-anvil experiments using 3D-printed “injection” molds is a low-cost and low failure-rate alternative for conducting reproducible experiments at high pressure in the multi-anvil apparatus.
more »
« less
- NSF-PAR ID:
- 10160889
- Date Published:
- Journal Name:
- Minerals
- Volume:
- 10
- Issue:
- 1
- ISSN:
- 2075-163X
- Page Range / eLocation ID:
- 4
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract The post-stishovite transition is a classic pseudo-proper typed ferroelastic transition with a symmetry-breaking spontaneous strain. This transition has been studied using high-pressure spontaneous strains, optic modes, and elastic moduli (Cij) based on the Landau modeling, but its atomistic information and structural distortion remain poorly understood. Here we have conducted synchrotron single-crystal X-ray diffraction measurements on stishovite crystals up to 75.3 GPa in a diamond-anvil cell. Analysis of the data reveals atomic positions, bond lengths, bond angles, and variations of SiO6 octahedra across the transition at high pressure. Our results show that the O coordinates split at ~51.4 GPa, where the apical and equatorial Si-O bond lengths cross over, the SiO6 octahedral distortion vanishes, and the SiO6 octahedra start to rotate about the c axis. Moreover, distortion mode analysis shows that an in-plane stretching distortion (GM1+ mode) occurs in the stishovite structure at high pressure while a rotational distortion (GM2+ mode) becomes dominant in the post-stishovite structure. These results are used to correlate with elastic moduli and Landau parameters (symmetry-breaking strain e1–e2 and order parameter Q) to provide atomistic insight into the ferroelastic transition. When the bond lengths of two Si-O bonds are equal due to the contribution from the GM1+ stretching mode, C11 converges with C12, and the shear wave VS1[110] polarizing along [110] and propagating along [110] vanishes. Values of e1–e2 and Q are proportional to the SiO6 rotation angle from the occurrence of the GM1+ rotational mode in the post-stishovite structure. Our results on the pseudo-proper type transition are also compared with that for the proper type in albite and improper type in CaSiO3 perovskite. The symmetry-breaking strain, in all these types of transitions, arises as the primary effect from the structural angle (such as SiO6 rotation or lattice constant angle) and its relevant distortion mode in the low-symmetry ferroelastic phase.more » « less
-
Obeid, I. ; Selesnik, I. ; Picone, J. (Ed.)The Neuronix high-performance computing cluster allows us to conduct extensive machine learning experiments on big data [1]. This heterogeneous cluster uses innovative scheduling technology, Slurm [2], that manages a network of CPUs and graphics processing units (GPUs). The GPU farm consists of a variety of processors ranging from low-end consumer grade devices such as the Nvidia GTX 970 to higher-end devices such as the GeForce RTX 2080. These GPUs are essential to our research since they allow extremely compute-intensive deep learning tasks to be executed on massive data resources such as the TUH EEG Corpus [2]. We use TensorFlow [3] as the core machine learning library for our deep learning systems, and routinely employ multiple GPUs to accelerate the training process. Reproducible results are essential to machine learning research. Reproducibility in this context means the ability to replicate an existing experiment – performance metrics such as error rates should be identical and floating-point calculations should match closely. Three examples of ways we typically expect an experiment to be replicable are: (1) The same job run on the same processor should produce the same results each time it is run. (2) A job run on a CPU and GPU should produce identical results. (3) A job should produce comparable results if the data is presented in a different order. System optimization requires an ability to directly compare error rates for algorithms evaluated under comparable operating conditions. However, it is a difficult task to exactly reproduce the results for large, complex deep learning systems that often require more than a trillion calculations per experiment [5]. This is a fairly well-known issue and one we will explore in this poster. Researchers must be able to replicate results on a specific data set to establish the integrity of an implementation. They can then use that implementation as a baseline for comparison purposes. A lack of reproducibility makes it very difficult to debug algorithms and validate changes to the system. Equally important, since many results in deep learning research are dependent on the order in which the system is exposed to the data, the specific processors used, and even the order in which those processors are accessed, it becomes a challenging problem to compare two algorithms since each system must be individually optimized for a specific data set or processor. This is extremely time-consuming for algorithm research in which a single run often taxes a computing environment to its limits. Well-known techniques such as cross-validation [5,6] can be used to mitigate these effects, but this is also computationally expensive. These issues are further compounded by the fact that most deep learning algorithms are susceptible to the way computational noise propagates through the system. GPUs are particularly notorious for this because, in a clustered environment, it becomes more difficult to control which processors are used at various points in time. Another equally frustrating issue is that upgrades to the deep learning package, such as the transition from TensorFlow v1.9 to v1.13, can also result in large fluctuations in error rates when re-running the same experiment. Since TensorFlow is constantly updating functions to support GPU use, maintaining an historical archive of experimental results that can be used to calibrate algorithm research is quite a challenge. This makes it very difficult to optimize the system or select the best configurations. The overall impact of all of these issues described above is significant as error rates can fluctuate by as much as 25% due to these types of computational issues. Cross-validation is one technique used to mitigate this, but that is expensive since you need to do multiple runs over the data, which further taxes a computing infrastructure already running at max capacity. GPUs are preferred when training a large network since these systems train at least two orders of magnitude faster than CPUs [7]. Large-scale experiments are simply not feasible without using GPUs. However, there is a tradeoff to gain this performance. Since all our GPUs use the NVIDIA CUDA® Deep Neural Network library (cuDNN) [8], a GPU-accelerated library of primitives for deep neural networks, it adds an element of randomness into the experiment. When a GPU is used to train a network in TensorFlow, it automatically searches for a cuDNN implementation. NVIDIA’s cuDNN implementation provides algorithms that increase the performance and help the model train quicker, but they are non-deterministic algorithms [9,10]. Since our networks have many complex layers, there is no easy way to avoid this randomness. Instead of comparing each epoch, we compare the average performance of the experiment because it gives us a hint of how our model is performing per experiment, and if the changes we make are efficient. In this poster, we will discuss a variety of issues related to reproducibility and introduce ways we mitigate these effects. For example, TensorFlow uses a random number generator (RNG) which is not seeded by default. TensorFlow determines the initialization point and how certain functions execute using the RNG. The solution for this is seeding all the necessary components before training the model. This forces TensorFlow to use the same initialization point and sets how certain layers work (e.g., dropout layers). However, seeding all the RNGs will not guarantee a controlled experiment. Other variables can affect the outcome of the experiment such as training using GPUs, allowing multi-threading on CPUs, using certain layers, etc. To mitigate our problems with reproducibility, we first make sure that the data is processed in the same order during training. Therefore, we save the data from the last experiment and to make sure the newer experiment follows the same order. If we allow the data to be shuffled, it can affect the performance due to how the model was exposed to the data. We also specify the float data type to be 32-bit since Python defaults to 64-bit. We try to avoid using 64-bit precision because the numbers produced by a GPU can vary significantly depending on the GPU architecture [11-13]. Controlling precision somewhat reduces differences due to computational noise even though technically it increases the amount of computational noise. We are currently developing more advanced techniques for preserving the efficiency of our training process while also maintaining the ability to reproduce models. In our poster presentation we will demonstrate these issues using some novel visualization tools, present several examples of the extent to which these issues influence research results on electroencephalography (EEG) and digital pathology experiments and introduce new ways to manage such computational issues.more » « less
-
null (Ed.)Electrochemical water splitting is one of the most promising approaches for sustainable energy conversion and storage toward a future hydrogen society. This demands durable and affordable electrocatalysts for the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER). In this study, we report the preparation of uniform Ni–P–O, Ni–S–O, and Ni–S–P–O electrocatalytic films on nickel foam (NF) substrates via flow cell-assisted electrodeposition. Remarkably, electrodeposition onto 12 cm 2 substrates was optimized by strategically varying critical parameters. The high quality and reproducibility of the materials is attributed to the use of a 3D-printed flow cell with a tailored design. Then, the as-fabricated electrodes were tested for overall water splitting in the same flow cell under alkaline conditions. The best-performing sample, NiSP/NF, required relatively low overpotentials of 93 mV for the HER and 259 mV for the OER to produce a current density of 10 mA cm −2 . Importantly, the electrodeposited films underwent oxidation into amorphous nickel (oxy)hydroxides and oxidized S and P species, improving both HER and OER performance. The superior electrocatalytic performance of the Ni–S–P–O films originates from the unique reconstruction process during the HER/OER. Furthermore, the overall water splitting test using the NiSP/NF couple required a low cell voltage of only 1.85 V to deliver a current density of 100 mA cm −2 . Overall, we demonstrate that high-quality electrocatalysts can be obtained using a simple and reproducible electrodeposition method in a robust 3D-printed flow cell.more » « less
-
more » « less
Abstract A simple and facile method was developed to fabricate functional bulk barium titanate (BaTiO3,
BT ) ceramics using the paste extrusion 3D printing technique. TheBT ceramic is a lead‐free ferroelectric material widely used for various applications in sensors, energy storage, and harvesting. There are several traditional methods (eg, tape casting) to process bulkBT ceramics but they have disadvantages such as difficult handing without shape deformation, demolding, complex geometric shapes, expansive molds, etc. In this research, we utilized the paste extrusion 3D printing technique to overcome the traditional issues and developed printable ceramic suspensions containingBT ceramic powder, polyvinylidene fluoride (PVDF ), N,N‐dimethylformamide (DMF ) through simple mixing method and chemical formulation. ThisPVDF solution erformed multiple roles of binder, plasticizer, and dispersant for excellent manufacturability while providing high volume percent and density of the final bulk ceramic. Based on empirical data, it was found that the maximum binder ratio with good viscosity and retention for desired geometry is 1:8.8, while the maximumBT content is 35.45 vol% (77.01 wt%) in order to achieve maximum density of 3.93 g/cm3(65.3%) for 3D printedBT ceramic. Among different sintering temperatures, it was observed that the sinteredBT ceramic at 1400°C had highest grain growth and tetragonality which affected high performing piezoelectric and dielectric properties, 200 pC/N and 4730 at 103 Hz respectively. This paste extrusion 3D printing technique and simple synthesis method for ceramic suspensions are expected to enable rapid massive production, customization, design flexibility of the bulk piezoelectric and dielectric devices for next generation technology.
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.3390/min10010004
