Pyrolyzed Ultrasharp Glassy Carbon Microneedles

Abstract

Polymeric microneedles fabricated via two-photon polymerization (2PP) lithography enable safe medical access to the inner ear. Herein, the material class for 2PP-lithography-based microneedles is expanded by pyrolyzing 2PP-fabricated polymeric microneedles, resulting in glassy carbon microneedles. During pyrolysis the microneedles shrink up to 81% while maintaining their complex shape when the exposed surface-area-to-volume ratio (SVR) is 0.025 < SVR < 0.04, for the temperature history protocol used herein. The derived glassy carbon is confirmed with energy-dispersive X-ray spectroscopy and Raman spectroscopy. The pyrolyzed glassy carbon has Young’s modulus 9.0 GPa. As a brittle material, the strength is stochastic. Using the two-parameter Weibull distribution, the glassy carbon has Weibull modulus of 3.1 and characteristic strength of 710 MPa. The viscoelastic response has characteristic time scale of about 10000 s. In vitro experiments demonstrate that the glassy carbon microneedles introduce controlled perforations across the guinea pig round window membrane (RWM) from the middle ear space into the inner ear, without damaging the microneedle. The resultant controlled perforation of RWM is known to enhance diffusion of therapeutics across the RWM in a predictable fashion. Hence, the glassy carbon microneedles can be deployed for mediating inner ear delivery.

Graphical Abstract

Microneedles have the potential to deliver therapeutics directly to the inner ear to treat hearing and balance disorders. In this study, glassy carbon microneedles are fabricated via Two-Photon Polymerization lithography followed by pyrolysis. The ultrasharp 22 μm diameter microneedles have high stiffness and strength and are able to introduce controlled perforations across the round window membrane into the inner ear.

1. Introduction

Hearing loss is the most common sensory impairment in humans, affecting more than 5% of individuals in industrialized nations.^[1] One of its major causes is inner ear disorder that is responsible for symptoms such as tinnitus, vertigo, and imbalance (e.g.,^[2]). The inner ear, which contains the cochlea, is embedded deep in the petrous part of the temporal bone—one of the hardest bones in the body. Three fluid-filled compartments coil up within the cochlea along with the organ of Corti. At the base of the cochlea are two orifices in the surface of the cochlear bone leading to the middle ear (e.g.,^[3]). One is the round window and the other one is the oval window, which are the only two nonosseous regions separating the middle ear chamber and the inner ear chamber. Both windows are covered by compliant membranes to separate the air-filled middle ear from the liquid-filled inner ear. However, the oval window membrane (OWM) is covered by the footplate of the stapes, which leaves the round window membrane (RWM) as the only potential nonosseous portal from the middle to the inner ear for therapeutic delivery and aspiration of perilymph fluid for diagnostic purposes. The human RWM spans approximately 2 mm with a thickness of about 70 μm^[4,5] and can only be accessed from within the middle ear space.

Due to its small size and inaccessibility, delivery of therapeutics into the cochlea across the delicate RWM carries significant risk of inducing permanent hearing loss. Therefore, therapeutic delivery to treat inner ear diseases typically^[6] occurs via three methods: 1) systemic injection that is accompanied by drawbacks such as the lack of clinical efficacy due to the presence of a blood-cochlea barrier and the potential for undesirable systemic side effects; 2) intratympanic injection of the therapeutic into the middle ear space to allow some of the therapeutic to diffuse across the RWM. This method avoids undesirable systemic side effects but the diffusion of the therapeutic across the RWM is limited and variable, and some molecules are not able to diffuse across the RWM.^[7-9] However, introduction of microperforations across the RWM significantly enhances both the rate and precision of delivery of therapeutic agents into the cochlea;^[10,11] and 3) direct intracochlear therapeutic delivery across the RWM, which minimizes potential systemic side effects and ensures that a precise volume of medication passes the RWM barrier.^[6,12,13]

Therapeutic delivery across the round window membrane with either of the latter two methods requires perforation of the RWM, with potential severe consequences to hearing and balance, often due to ripping of the RWM (e.g.,^[14]). In addition, as the cochlea is a closed constant-volume chamber, intracochlear injection of relatively large volumes into the inner ear may lead to hearing loss. Thus, while RWM-perforation-mediated diffusive intratympanic delivery may be less precise than intracochlear delivery, diffusion-mediated delivery may be preferable for many therapeutic agents.

A potential concern of introducing perforations across the RWM is leakage of perilymph from the cochlea into the middle ear. The possibility of perilymph leakage can be mitigated by reducing the perforation size, which, in turn, is determined by the size of needle. Decreasing the diameter of a perforation prevents leakage of perilymph without slowing the diffusive transport of therapeutic reagents.^[10]

The glassy carbon microneedles reported herein are intended to perforate the RWM to enhance diffusion-mediated delivery of therapeutics into the cochlea after intratympanic delivery into the middle ear space. Based on a detailed understanding of the geometry and mechanical properties of the RWM,^[15] our group has designed, fabricated, and deployed a suite of polymeric microneedles and metallic microneedles that can safely perforate the RWM without ripping or tearing. The polymeric microneedles designed by our group^[16,17] for RWM perforation were fabricated using two-photon polymerization lithography, which is akin to a 3D printing technology. Two-photon polymerization lithography (2PP) can produce mesoscale to nanoscale structures of nearly any 3D shape with submicrometer resolution in a wide range of polymeric materials (e.g.,^[18]). The method is based on two-photon polymerization with ultrashort laser pulses: when focused into the volume of a photosensitive material (or photoresist), the pulses initiate 2PP via two-photon absorption and subsequent polymerization.^[19] The 2PP method achieves a resolution beyond the diffraction limit by choosing an appropriate laser-pulse energy and exposure time to activate nonlinear photochemical reactions that initiate or propagate only if the exposure energy is greater than a critical value.^[20] The polymer resulting from 2PP of IP-S photoresist^[18] has an elastic stiffness of 5.1 GPa and hardness of 220 MPa based upon Berkovich indentation, which has adequate strength to perforate the RWM, and thus is a good candidate for fabricating microneedles to perforate the RWM.

Our group has demonstrated 2PP-fabricated microneedles—with a 100 μm shaft diameter and a 500 nm tip radius of curvature—that can perforate the guinea pig (GP) RWM in vitro with a mean force of 1019 mN.^[16] Importantly, the ultrasharp microneedle separated—rather than cut—the collagen and elastic fibers within the RWM during the perforation process. Subsequent studies by^[21] showed that the GP RWM healed readily with no hearing damage, demonstrating that the microneedle introduces perforations in vivo with a minimal degree of trauma. Significantly, similar 2PP-printed microneedles can perforate cadaveric human RWM tissue by the same “separation not scission” mechanism.^[17] Further, we designed and fabricated hollow 100 μm diameter microneedles and aspirated 1 μL of perilymph across the GP RMW that was analyzed using proteomics,^[13,22] again with no functional or structural consequence. Finally, after introducing microscale perforations across a GP RWM as well as across a proxy artificial membrane, the resulting perforations significantly enhance the therapeutic diffusion rate across the RWM.^[10,11]

In addition to polymeric microneedles, our group developed the two-photon templated electrodeposition (2PTE) methodology to fabricate fully metallic microneedles^[23,24] that can safely perforate the RWM. Essentially, 2PP photolithography is used to fabricate a mold containing a cavity with the shape of the microneedle, followed by electrodeposition of metal into the mold, after which the mold is chemically dissolved to release the microneedle.

In the present study, we further expand the classes of materials from which we make microneedles by applying pyrolysis to the 2PP polymeric microneedles. Pyrolysis converts carbon-rich polymers to functional high-carbon solids at elevated temperatures (400 to 1800 °C) under an inert atmosphere. During pyrolysis, organic molecules in a carbon-rich polymer break down and the resulting gases and volatile products leave the sample and nonvolatile residues form large disordered molecules that typically are richer in carbon than the precursor.^[25]

The structure glassy carbon is intermediate between the 3D crystalline structure of graphite and the “random layer structure” of the nongraphitic carbons.^[26] It has favorable properties including high-temperature resistance, high hardness, low density, low electrical resistance, low friction, and low thermal resistance. However, these physical, mechanical, electrical, and chemical properties strongly depend on the heat treatment protocol,^[27-29] which presents the opportunity to tailor the properties of the pyrolyzed material. Importantly, pyrolyzed materials can be biocompatible^[27,30] and as such have been widely used in medical devices.^[31,32]

SU-8 photoresist is one of the polymers that has long been used as a precursor material for pyrolysis. Many protocols have been developed to transform SU-8 precursors into carbon electrodes or transdermal microneedles.^[33-36] However, our group uses carbon-rich IP-S photoresist instead as the precursor because it is optimized for use with the 2PP equipment, and thus it provides us with the design freedom to create relatively large and complicated 3D shapes.

The pyrolysis protocol developed by^[36] for IP-S serves as the starting point for our study. Other studies on carbonization of 2PP-fabricated IP-S precursors have been conducted due to its rising popularity in photolithography.^[37-39] The structures explored in those studies are mainly electrodes with conical and spherical shapes with relatively simple geometries compared to microneedles that must be mounted in a surgical tool. In addition, the structures from the other studies are significantly smaller than the microneedles in our study.

During pyrolysis, the carbon-rich polymer precursor shrinks significantly while generally retaining its overall shape, which creates the opportunity to fabricate very small structures. However, pyrolysis occurs most rapidly on the surface of the polymer which leads to the development of surface tensile stresses that can induce fracture.^[40] In addition, nonspatially-uniform pyrolysis—especially of complex shapes—can lead to significant changes in shape and substrate delamination.^[41] Thus, isotropic pyrolysis of complex 2PP-fabricated IP-S precursors remains challenging.

Here, we propose a solution to this challenge, in which a protocol is developed to fabricate pyrolyzed microneedles from an IP-S precursor. The resulting microneedles have high design freedom and expand the material category of 2PP-based microneedles, and can be fabricated in a reproducible manner. We demonstrate the utility of the resulting fabricated glassy carbon microneedles by perforating in vitro the RWM of a guinea pig model to confirm that they are sharp and robust enough to make microperforations across RWM. The resulting microneedles have a significantly higher stiffness and hardness than their nonpyrolyzed counterparts and also are significantly smaller.

In what follows, Section 2 describes the general design of the microneedle. Section 3 introduces the fabrication method of the pyrolyzed microneedles, methods to characterize the microneedles’ material, and the protocol to perform in vitro perforation experiments. The results are reported in Section 4. Important design factors, future improvements, and applications are discussed in Section 5. Conclusions are presented in Section 6.

2. Microneedle Design

A prototypical solid microneedle design from our group is shown in Figure 1. It starts from a tip of radius R_t and tapers at an angle α to a constant shaft diameter D_o, with a taper-plus-shaft length of L and a base with maximum diameter D_b designed to fit into the lumen with diameter, D_L, of a commercially available blunt stainless-steel hollow needle. A Luer lock is affixed to the other end of the stainless-steel needle. Table 1 lists the dimensions, angles, and materials of the microneedle suite developed previously by our group.

Figure 1.

Microneedle designed to mount on blunt stainless-steel hollow needle. The microneedle consists of a base indicated by dimensions D_b and D_L as well as the microneedle itself consisting of the shaft with outside diameter D_o and the tapered portion.

Table 1.

Design parameters of microneedle suite. Microneedle in final row is hollow with lumen diameter denoted D_i.

Material	Rt [μm]	α [deg]	L [μm]	Do [μm]	Di [μm]	Db [μm]	DL [μm]	References
IP-S	0.5	18	200	100	N/A	320	170	[16]
IP-S	5	60	350	150	N/A	300	160	[17]
Gold-coated copper	7.5	10	600	100	N/A	600	280	[23]
IP-S	0.5	24	435	100	35	310	139	[22]

The 2PP-fabricated microneedles prior to pyrolysis have significantly larger dimensions than after pyrolysis. The dimensions prior to pyrolysis are R_t = 7.5 μm, D_o = 100 to 120 μm, μ = 10° to 24°, L = 600 to 850 μm, D_b = 600 μm, and D_L = 280 μm. The dimensions of the pyrolyzed microneedle depend on the pyrolysis parameters and are listed in Table 2.

Table 2.

Dimensions of microneedles before and after pyrolysis.

IP-S microneedles		Glassy carbon microneedles		Shrinkage factor
Do [μm]	L [μm]	Do [μm]	L [μm]	Diameter	Length
100	850	25	195	0.75	0.77
120	600	32	120	0.73	0.80
150	600	48	140	0.68	0.77

3. Methods

The process to fabricate the pyrolyzed microneedles is now described. The design was effected using the computer-aided design (CAD) software SolidWorks (Dassault Systems SolidWorks Corporation, USA) as illustrated in Figure 2 and exported as stereolithography (STL) files as shown in Figure 3a.

Figure 2.

Images of the design of microneedle polymeric precursor: a) isometric-sectional view; b) cross-sectional view showing tapered shaft, plate, hole in the base and four springs.

Figure 3.

Process flow for the fabrication of pyrolyzed microneedles: a) design and export as STL file; b) process file; c) 2PP-fabricated polymeric microneedles; d) clean uncured photoresist; e) pyrolyze polymeric microneedles; f) mount pyrolyzed microneedles. The syringe blunt needle model is from GrabCAD.^[53]

3.1. Fabrication of Polymeric Microneedles

Figure 3a-d illustrates the fabrication process of 3D printed polymeric microneedles. The STL file was first imported into software DeScribe^[18] to process the CAD model for 2PP fabrication. Processing parameters were selected with fixed slicing distance of 0.5 μm while employing solid fill mode, hexagonal splitting mode, and lexical block order. A single crystal silicon (100) wafer served as substrate and was rinsed with acetone and isopropyl alcohol (IPA). IP-S photoresist was drop cast onto the substrate. Then a volume within the IP-S corresponding to the desired microneedle was polymerized via exposure to 2PP using the Photonic Professional GT system.^[18] The equipment has X,Y resolution of 400 nm in the plane of the substrate, vertical resolution of 1000 nm, maximum X and Y dimensions of 100 mm, and maximum height of 8 mm. After exposure to the 2PP process, the polymerized structures were developed in 1-methoxy-2-propanol acetate for 20 min and subsequently immersed in IPA bath for 5 min.

3.2. Pyrolysis

The samples were pyrolyzed in a Thermo Scientific Lindberg/Blue M TF55030A tube furnace (Hogentogler, USA). The pyrolysis temperature is plotted versus time in Figure 3e. A gas consisting of 5% oxygen and 95% argon at 101 kPa pressure flowed at 40 cm³ min⁻¹ while ramping the temperature up from ambient to 450 °C at 0.5 °C min⁻¹, holding at 450 °C for 30 min, followed by ramping down to room temperature at 10 °C min⁻¹. The flowing gas was then changed to 5% hydrogen and 95% argon at 101 kPa at a flow rate of 40 cm³ min⁻¹, while the temperature ramped up to 900 °C at 1 °C min⁻¹, holding at 900 °C for 30 min, and finally ramping down to room temperature at 10 °C min⁻¹. The rationale for choosing these pyrolysis steps is discussed in Section 5.

3.3. Carbon Microneedles Mounting

The pyrolyzed microneedles were caught up individually with tweezers and manually inserted onto a Gauge 32 syringe tip (sterile standard blunt needle) and adhered with epoxy as shown in Figure 3f. Honey was used on the tweezers’ tip to ease the manipulation resulting from the microneedle’s tiny size. The commercially available syringe tip was chosen as the mounting base because clinicians are familiar with this tool. Finally, the honey and other residues on the microneedle were cleaned with IPA.

3.4. Microneedle Characterization

Scanning electron microscopy (SEM, Zeiss, Germany) was used to characterize the pyrolyzed microneedles’ geometry.The material composition characterization of the microneedles before and after pyrolysis was performed by energy-dispersive X-ray spectroscopy (EDS, Bruker, UK). Raman spectroscopy analysis (Renishaw, UK) was utilized to determine crystallinity.

To characterize the mechanical properties of pyrolyzed carbon, micropillars were 2PP-fabricated and pyrolyzed using the same protocol as for the microneedles. The pyrolyzed micropillars had a height of 70 μm with diameter of 22 μm (measured at midheight) and remained adhered to the silicon substrate. The 22 μm diameter was chosen to match the microneedles’ shaft diameter. A G200 Nanoindenter (KLA, USA) was used to load the top of each micropillar using a 1 mm diameter flat punch.

To characterize the pyrolyzed material’s viscoelasticity, a creep test was executed by prescribing a 100 mN constant force on the pyrolyzed micropillars for 10000s. To measure Young’s modulus, three micropillars were compressed to a maximum compressive force of 150 mN at a 0.01 s⁻¹ strain rate and another three were compressed to 200 mN at the same strain rate. The force was held constant for 10s and then unloaded at 5 mN s⁻¹. For compressive strength measurements, the indenter tip translated the top of each of 11 micropillars downward by 25 μm at a 0.001 s⁻¹ strain rate and the force and displacement at rupture were recorded.

The electrical resistance of 5 pyrolyzed micropillars was measured with VERSA Modular Probe System^[42] at room temperature. One probe tip (7B) was placed in contact with the top of a pyrolyzed micropillar and the probe end was connected to positive of Keithley 2400 Sourcemeter. The other probe tip (also 7B) was placed in contact with the bottom of the pyrolyzed micropillar and the probe end was connected to ground of the sourcemeter. Electric potential was supplied by the sourcemeter from 0 to 1 V in increments of 0.05 V and the resulting current was measured. Electrical resistance of each pyrolyzed micropillar was calculated according to Ohm’s law.

3.5. Round Window Perforation with Microneedle

The mounted pyrolyzed microneedles were used to perform in vitro perforation of a guinea pig RWM using the same protocol as the perforation tests by^[16] with 2PP-fabricated microneedles. All procedures were approved by the Columbia University Institutional Animal Care and Use Committee (IACUC) with protocol number AC-AABA5450. Hartley strain male guinea pigs were euthanized and their intact temporal bones were harvested followed by immediate blunt dissection. An Osada Electric Handpiece System (Osada, Inc., USA) was utilized to drill and remove the bone surrounding the RWM to obtain an unobstructed view. The sample was then rinsed with 0.9% saline solution and kept refrigerated in the solution (up to a maximum of 24 h) until the perforation test. During the experiment, drops of sterile 0.9 % saline solution were applied at regular intervals to keep the membrane from drying. We used a special-built microindenter^[16] to introduce the perforations. The microindenter has a motorized stage for manipulating the harvested RWM’s position as well as a motorized linear translator to control the needle. A force transducer (98 mN full scale range) measured the axial force exerted on the microneedle during indentation. The resulting perforations were imaged with confocal laser scanning microscopy (Zeiss, Germany) and the microneedles after perforation were imaged with SEM to determine damage.

3.6. Statistical Analysis

Averages and standard deviations were used for microneedle dimensions after pyrolysis. The 95% intervals^[43] were calculated for Young’s modulus, Weibull modulus, characteristic strength, and electrical properties. Statistical analyses were carried out using MATLAB.

4. Results

4.1. Glassy Carbon Microneedle Dimensions

The pyrolysis protocol enables consistent fabrication of pyrolyzed microneedles. Figure 4a shows an IP-S polymer microneedle prior to pyrolysis. The base of the microneedle is attached to four polymeric springs that were fabricated along with the microneedle and are in turn attached to the substrate. Both the microneedle and the springs shrink during pyrolysis. The springs decouple the microneedle from the substrate, preventing damage to the microneedle during pyrolysis. Figure 4b shows a representative pyrolyzed microneedle mounted in the hollow stainless steel blunt needle. Figure 4c shows the tip of an as-fabricated pyrolyzed microneedle.

Figure 4.

Images of a representative pyrolysis result: a) optical image of a 3D-printed unpyrolyzed microneedle of diameter D_o = 100 μm; b) SEM image of the corresponding pyrolyzed microneedle with diameter D_o = 24 ± 2 μm (n = 5) mounted on Gauge 32 stainless steel syringe tip;c)SEM image of the enlarged view of an as-fabricated pyrolyzed microneedle tip.

Prior to pyrolysis the microneedles of Figure 4a have outside diameter D_o = 100 μm, length L = 600 μm, and tip radius of R_t = 7.5 μm. After pyrolysis, the outside diameters have decreased to D_o = 24 ± 2 μm (n = 5), the length to L = 116 ± 10 μm (n = 5), and tip radius to R_t = 1 μm, achieving shrinkage factors of 0.76 ± 0.02 and 0.81 ± 0.02 for diameter and length, respectively. The shrinkage factor (SF) is calculated as SF = (ℓ_p – ℓ_c)/ℓ_p, where ℓ_p is the dimension of polymeric microneedle and ℓ_c is the dimension of the pyrolyzed carbon microneedle.

Microneedles with other dimensions were also fabricated and pyrolyzed to determine the protocol’s robustness. The microneedles with larger initial diameters tend to shrink less than the smaller ones. The dimensions before and after pyrolysis are reported in Table 2.

4.2. Material Characterization

The EDS results of material composition are given in Table 3. Nitrogen was not found in the pyrolyzed microneedle, indicating the nitrogen component is removed during pyrolysis. Oxygen atomic percentage decreases significantly from 19.69% to 3.85%, showing that only a small amount of oxygen remains. The pyrolyzed microneedle is mainly composed of carbon with an atomic percentage of 94.22%. A silicon component is detected that is assumed to result from the silicon substrate.

Table 3.

EDS analysis results for IP-S and pyrolyzed carbon microneedles.

Atomic [%]	IP-S microneedle	Glassy carbon microneedle
Carbon	70.70	94.22
Oxygen	19.69	3.85
Nitrogen	1.64	–
Silicon	0.06	1.77

The results of Raman spectroscopy analysis in Figure 5 show a peak at about 1350 cm⁻¹ and another peak at about 1580 cm⁻¹. The first peak is around the D-band value (1357 cm⁻¹) and the second peak matches the G-band value (1580 cm⁻¹) that is related to sp2-hybridized carbon bonds. Well-crystallized graphite with a small particle size shows a D-band in addition to the G-band.^[44] The intensity ratio of D-band over G-band (I_D/I_G) is around 0.93, which is within glassy carbon’s intensity ratio (0.91–1.32) reported in the literature.^[36,44] Overall, the EDS and Raman spectroscopy results verify that the resulting material is glassy carbon.

Figure 5.

Raman spectroscopy analysis result for the pyrolyzed carbon microneedle.

Figure 6 shows micropillars fabricated using the same protocol used to make the microneedles. An array of micropillars was distributed spatially on the substrate, as shown in Figure 6a, ensure that all micropillars underwent the same fabrication and pyrolysis protocol. Figure 6b shows a higher magnification image of one microneedle; the dark ring around the base of the microneedle indicates the diameter of the 2PP-fabricated pillar prior to pyrolysis.

Figure 6.

Micropillars used for mechanical characterization: a) SEM image of a representative pyrolyzed micropillars array on a silicon substrate; b) SEM image of the enlarged tilted view of a micropillar (diameter = 22 μm, height = 70 μm).

The micropillars were fabricated directly on the substrate to allow subsequent mechanical characterization. Springs connecting the base of the micropillar to the substrate were not necessary because the diameter of the micropillar is significantly less than D_L in Figure 1. Prior to pyrolysis, each 2PP-fabricated micropillar was a right cylinder with diameter 100 μm and height of 400 μm, with a flat top. After pyrolysis, the diameter was 22 μm and height of 70 μm with shrink factors, respectively, of 0.78 and 0.83. The top of the micropillars developed a curvature during pyrolysis, as in Figure 6b, due to anisotropic shrinkage.

Glassy carbon is viscoelastic, so its mechanical response is a function of the rate and the time scale of loading. Here, we characterize the material properties over short time intervals of a few seconds as well as over long time periods of several hours.

The short-time scale mechanical properties dictate the microneedle performance during the process of perforating a RWM. A nanoindenter (KLA G200) with a 1 mm diameter cylindrical tip was used to deform the micropillars in compression axially at a 0.1 s⁻¹ strain rate to a force of either 150 mN (n = 3) or 200 mN (n = 3) after which the load was removed at 5 mN s⁻¹. A representative engineering stress versus engineering strain curve is shown in Figure 7a. The initial nonlinear response is due to the postpyrolysis curvature of the top of the micropillar shown in Figure 6b. Hence, the Young’s modulus is found by calculating the slope of the first 5–25% of the unloading curve. The mean Young’s modulus is E = 9.0 GPa with 95% confidence interval (CI) from 8.8 to 9.2 GPa (n = 6).

Figure 7.

Short-term mechanical response: a) representative stress versus strain response to measure Young’s modulus of the material during unloading; b) representative stress versus strain curve for compressive strength; c) Weibull probability plot for the glassy carbon material; d) SEM image of a crushed micropillar after the compressive strength test.

To characterize the strength over short-time scale loading, the micropillars were compressed axially at a 0.01 s⁻¹ strain rate until they suddenly shattered into pieces; the engineering stress versus engineering strain response is shown in Figure 7b, demonstrating catastrophic brittle fracture. The strength of brittle materials is stochastic and depends on the distribution and sizes of flaws in the micropillars. Thus, a two-parameter Weibull distribution model with maximum likelihood estimator is applied to fit the compressive strength data as shown in Figure 7c.

The two-parameter Weibull distribution^[45] is taken as

$$P_{\mathrm{f}}=1-\text{exp}\left[-{\left(\frac{\sigma }{{\sigma }_{\theta }}\right)}^{\mathrm{m}}\right]$$ (1)

where P_f is probability of failure, σ is the measured uniaxial compressive strength, m is the Weibull modulus, and σ_θ is Weibull characteristic strength. The parameters σ_θ and m are determined as maximum likelihood estimates^[46] from

$${\widehat{\sigma }}_{\theta }={\left[\frac{1}{n}\sum _{i=1}^n{\sigma }_i^{\widehat{\mathrm{m}}}\right]}^{\frac{1}{\widehat{\mathrm{m}}}}$$ (2)

$$\widehat{m}=\frac{n}{\frac{1}{{\widehat{\sigma }}_{\theta }}{\sum }_{i=1}^n{\sigma }_i^{\widehat{\mathrm{m}}}\log {\sigma }_i-{\sum }_{i=1}^n\log {\sigma }_i}$$ (3)

where ${\widehat{\sigma }}_{\theta }$ and $\widehat{m}$ are unbiased estimators of the parameters σ_θ and m. As m and σ_θ are positive parameters, the logarithmic transformation is used to obtain the 95% CI.^[43]

The Weibull modulus is m = 3.1 with 95% CI from 1.9 to 4.9 and characteristic strength σ_θ = 710 MPa with 95% CI from 580 to 870 MPa as shown in Figure 7c. The measured highest load of a micropillar was 358 mN that corresponds to a compressive strength of 990 MPa.

To characterize the mechanical response over long time scales, viscoelastic creep tests were performed on a set of pyrolyzed micropillars shown in Figure 6. The pyrolyzed micropillars were loaded with a 1 mm cylindrical tip with a nanoindenter (KLA G200) to a compressive loading of 100 mN (corresponding to a 263 MPa compressive stress) at a 0.01 s⁻¹ strain rate, after which the force was maintained at a constant load of 100 mN for 10000s during which time the micropillar experienced creep strain. Figure 8a shows representative results of a creep test.

Figure 8.

Long-term mechanical response: a) experimental creep test results, along with comparison between fitting and experimental data; b) discrete retardation spectrum of the material compliance Prony series, where the instantaneous creep compliance is J₀ = 9.96 × 10⁻⁵ Mpa⁻¹.

The viscoelastic creep tests were analyzed assuming the micropillar to be load-free prior to time t = 0, at which time a linear ramp up stress σ(t) is applied until t = t₁. The strain ε(t) can be expressed with the hereditary integral

$$\epsilon (t)={\int }_0^{\mathrm{t}}J(t-\xi )\frac{\mathrm{d}\sigma (\xi )}{\mathrm{d}\xi }\mathrm{d}\xi$$ (4)

where dσ(ξ)/dξ is the stress rate and J(t) is the compliance of the material defined as

$$J(t)=J_0+J_0\sum _{i=1}^n{p}_i(1-e^{-\mathrm{t}∕{\tau }_i})$$ (5)

where J₀ is the instantaneous creep compliance of the material, p_i is the i^th Prony constant, and τ_i is the i^th Prony retardation time constant. For t₁ < t ≤ t₂, during which a constant stress is prescribed, the hereditary integral can be expressed as

$$\epsilon (t)={\int }_0^{{\mathrm{t}}_1}J(t-\xi )\frac{\mathrm{d}\sigma (\xi )}{\mathrm{d}\xi }\mathrm{d}\xi +{\int }_{{\mathrm{t}}_1}^{\mathrm{t}}J(t-\xi )\frac{\mathrm{d}\sigma (\xi )}{\mathrm{d}\xi }\mathrm{d}\xi$$ (6)

where the second term is zero.

Characterization of the material behavior requires finding the parameters τ_i and p_i to make Equation (6) provide a good fit over all time of the creep test data.^[47] While many methods exist to determine τ_i and p_i, here a sign control method is used.^[48] First, 15 values of τ_i were chosen evenly spaced in log time over 8 decades. Once these relaxation times were selected, the coefficients p_i were found with MATLAB (Mathworks, USA) via nonlinear optimization to minimize the difference between fitting and experimental data with p_i constrained as positive.

Figure 8b shows the Prony series representation of the material. The viscoelastic response is well captured with this method by comparing the experiment and fitting results in Figure 8a. The dominant time constant is about 10000s.

The electrical resistance of five micropillars was measured. The 95% CI for micropillar electrical resistance is 29.7 ± 5.4 kΩ (n = 5). Given the uniform cross-section and uniform flow of electric current in the micropillar, the electrical conductivity σ of the glassy carbon is calculated as

$$\sigma =\frac{l}{RA}$$ (7)

where R is the electrical resistance of a uniform specimen of the material, A is cross-sectional area, and l is the length of the specimen. Glassy carbon produced by our protocol has an electrical conductivity with a 95% confidence interval of 9.8 ± 1.2 Ω⁻¹ (n = 5).

4.3. Perforation Test

In order to assess the performance of the glassy carbon microneedles perforating the guinea pig RWM, microperforation experiments were performed with (n = 5) pyrolyzed microneedles. One of the perforations is demonstrated in Figure 9. The resulting perforations are oval shaped as in Figure 9b. The microneedle tip was undamaged by the perforation process as in Figure 9c. As shown in Figure 9d, the force exerted on the microneedles first increases and reaches a peak as the microneedle tip perforates through the RWM. The force then decreases until the shaft of the microneedle has fully penetrated the RWM. The maximum forces experienced are in the range 1.1 to 2.3 mN as shown Figure 9e. Just as for the 2PP-fabricated polymeric microneedles,^[16] the pyrolyzed microneedles are able to penetrate the RWM successfully without ripping or tearing the RWM. Earlier our group^[13,21,22] investigated the anatomical and functional consequences of microperforations introduced by 100 μm diameter polymeric microneedles through the guinea pig RWM in vivo. The studies conclude that the microperforations can heal within 72 h without causing significant anatomic or physiologic dysfunction. Thus, the glassy carbon microneedles (with significantly smaller diameter) are likewise expected to perforate the RWM without consequent hearing loss.

Figure 9.

In vitro perforation experiment results: a) optical image illustrating the perforation experiment setup; b) confocal image of a perforated RWM (perforation indicated by yellow arrow); c) SEM image of the microneedle’s tip after perforation; d) force versus time plot for a representative perforation experiment; e) maximum force measured during each perforation tests. The bottom and top edges of blue box indicate the 25th and 75th percentiles, respectively. The horizontal red central mark indicates the median. The vertical whiskers extend to the highest and lowest data points.

5. Discussion

A significant challenge in this project was to avoid rupture of the polymeric material due to the buildup of high-pressure gas generated by noncarbon atoms that are broken down into compounds during pyrolysis. The key to avoiding rupture is to allow these gases to diffuse out of the polymer at a rate faster than they are generated. Decreasing the peak temperature and reducing the temperature ramp rate during pyrolysis decreases the rate at which gases are generated. Increasing the structure’s surface-area-to-volume ratio (SVR) shortens the average diffusion length for a gas molecule to escape a surface exposed to the environment and thus increases the rate at which gas diffuses out of the material. We employed both strategies.

Regarding the temperature, we used the same peak temperatures of 450 °C and 900 °C as in the protocol by^[35] but reduced the temperature ramp rate from 5 °C min⁻¹ to 0.5 °C min⁻¹.

Regarding SVR, the complex shape of the microneedles complicates the ability to increase the SVR. In Figure 1, the SVR of the microneedle shaft-plus-taper portion and the SVR of the base portion of the structure must have similar values, to ensure that they have similar shrink rates and maintain the same general shapes after pyrolysis. During the design process, the SVR is calculated individually for the shaft-plus-taper portion and the base portion, where the surface area calculation includes only surfaces that are exposed to the environment.

However, in the prototypical microneedle design of Figure 1, the dimensions of the base portion necessary to be mounted in the hollow stainless-steel needle render the SVR significantly lower than the SVR of the shaft-plus-taper portion. We address this in two ways. First the microneedles were 2PP-fabricated on polymeric springs that were also 2PP-fabricated. This exposed the bottom surface of the base to the environment and increased the SVR. Second, we designed the base portion to be partially hollow as shown in Figure 2b. The interior diameter and depth of the hole in the base are design parameters, and the SVR increases as the interior diameter increases, and also as the depth increases.

Our results show that having sufficient nearby surface area for degassing requires SVR > 0.025 for the reported temperatures and ramp rates. From Figure 1, the small ratio D_o/L in for the shaft-plus-taper portion easily maintains SVR > 0.025, but the base portion required a concentric hole of a minimum diameter and depth to meet that criterion.

Moreover, when maintaining a high SVR for degassing, the base and plate should not be too thin otherwise they will not sustain the residual stresses induced during shrinkage or the applied stresses during usage. Our results show that the SVR needs to be less than 0.04 to maintain structural integrity during pyrolysis, which places an upper bound on the dimensions of the concentric hole in the base.

Anisotropic shrinkage during pyrolysis due to the complex shape of the microneedle introduces stress concentrations. Therefore, we included fillets at the transitions between geometry features to decrease stress concentrations.

The very significant shrinkage during pyrolysis led to mechanical failure of the microneedle base during pyrolysis when the precursor microneedle was 2PP-fabricated directly on a silicon substrate. Inspired by,^[49] we fabricated the precursor microneedle atop polymeric springs rather than directly on a substrate—as shown in Figure 4a—to decouple the microneedle base from the substrate. The springs pyrolyze as well during pyrolysis but were not required to maintain structural integrity.

In order to successfully perforate the RWM, the microneedle’s tip radius of curvature must be significantly smaller than the guinea pig RWM’s thickness, which ranges 10 μm to 30 μm. Indeed, in ref. [16], the radius of the microneedle tip is 500 mm. However, due to anisotropic pyrolysis as a consequence of rapid spatial variations in SVR near the microneedle tip, the shape near the tip is severely distorted during pyrolysis. Therefore, we used an as-fabricated tip radius of curvature of 7.5 μm, which shrank relatively uniformly to a tip radius of 1 μm as shown in Figure 4.

The Young’s modulus (8.8 to 9.2 GPa) and strength (580 to 870 MPa) of the glassy carbon from our pyrolysis protocol are smaller than those reported elsewhere. For example,^[49,50] reported Young’s modulus values ranging from 15 to 47 GPa and strength values from 1.2 to 7 GPa. One possible reason for this is that the micropillars tested in this work are significantly larger than the pyrolyzed structures reported elsewhere.^[49,50] Therefore, our pyrolyzed structures have a smaller SVR than in the other structures, which may lead to less complete pyrolysis. In addition, the larger structures can have larger defects that lead to a reduction of the stochastic strength. By pyrolyzing the microneedles in a vacuum instead of flowing inert gas and further exploring the 2PP fabrication parameters including contour counts and slicing distances, a glassy carbon material with a lower defect concentration can possibly be achieved.

Electrical conductivity of glassy carbon can span several orders of magnitude as influenced by heating temperature, dwell time, and sample size (e.g.^[28,29]). For example, poly(furfuryl alcohol) resin-derived glassy carbon has electrical conductivity ranging from 0.1 Ω⁻¹ m⁻¹ when carbonized at 600 °C, to 2 × 10⁴ Ω⁻¹ m⁻¹ when carbonized at 2000 °C;^[51] this glassy carbon is mostly in wire or thin film form. Electrical conductivity (8.6 to 10.9 Ω⁻¹ m⁻¹) of our glassy carbon falls in the lower part of the range. It is known that electrical conductivity of glassy carbon increases with the annealing temperature.^[52] The sp² bond facilitates electron transport whereas diamond-like sp³ bonding induces high resistivity. The I_D/I_G Raman spectroscopy ratio which characterizes the sp²/sp³ ratio of our glassy carbon is around 0.93, as shown in Figure 5. This ratio increases with annealing temperature and the structure can approach 100% sp²-bonded when the heating temperature is over 2000 °C.^[28] As a result, our glassy carbon’s electrical conductivity can be potentially increased by further annealing to a higher temperature.

Our group demonstrated previously that perforation of the guinea pig RWM in vitro resulted in a 35× increase in the rate of delivery of a molecular agent in saline solution across the RWM as compared to a RWM without perforations.^[10] The microperforations in that study had an average radius of 23 μm, with a relatively large standard deviation in radius due to the imprecision of the microneedles. The glassy carbon microneedles reported herein are significantly sharper and have more uniform dimensions than in that study. Therefore, as discussed by,^[10] the smaller perforations will reduce leakage of perilymph from the cochlea to the middle ear without reducing the rate of diffusive flow into the cochlea. Finally, our group showed that the rate of delivery increases linearly with the total perforation area across a multiply perforated proxy membrane.^[11]

6. Conclusion

In conclusion, the fabrication method presented here introduces a consistent protocol to pyrolyze two-photon photolithography (2PP)-fabricated polymeric microneedles into ultrasharp (tip radius of 1 μm) glassy carbon microneedles that are smaller, stiffer, and more biocompatible than the polymeric ones. We found that the exposed SVR of each portion of the 2PP-fabricated microneedle structure should be within the range 0.025 < SVR < 0.04 to avoid rupture and to maintain structural integrity during the pyrolysis process. The microneedles shrink by a factor of up to 81% during pyrolysis. EDS and Raman analyses confirm that the remaining material after pyrolysis is glassy carbon. It contains 94.22% carbon and shows Raman spectroscopy D-band and G-band at 1350 and 1580 cm⁻¹, respectively.

The short-term mechanical response (over a characteristic time of a few seconds) of the glassy carbon is elastic with a brittle failure mechanism. Compression tests on the 22 μm diameter micropillars show that the pyrolytic carbon has Young’s modulus of 9.0 GPa with 95% confidence interval from 8.8 to 9.2 GPa (n = 6), and fails in catastrophic brittle manner. Its measured compressive strength ranges from 280 to 990 MPa, which as a brittle material depends upon the flaw sizes and distribution. The measured compressive strength values are fitted to the two-parameter Weibull distribution, finding a Weibull modulus of m = 3.1 with 95% CI ranging from 1.9 to 4.9 and a characteristic strength σ_θ = 710 MPa with 95% CI ranging from 580 to 870 MPa.

The long-term mechanical response of the glassy carbon is viscoelastic. A discrete spectrum of retardation times for the material is determined by fitting the creep test result to the generalized Kelvin model expressed as a Prony series. The dominant viscoelastic time constant is about 10000s.

Probe station measurements of 22 μm diameter micropillars yielded electrical resistance with a 95% confidence interval ranging from 24.3 to 35.1 Ω m. With uniform cross section and uniform current assumptions, the pyrolyzed glassy carbon 95% CI of electrical conductivity can be calculated from pyrolyzed micropillars electrical resistance as 9.8 ± 1.2 Ω⁻¹ m⁻¹ (n = 5).

The pyrolyzed microneedles can be manually mounted on a Gauge 32 syringe tip and utilized to perforate in vitro the guinea pig round window membrane (RWM). The perforation tests illustrate that the glassy carbon microneedles are sharp and robust enough to introduce lens-shaped holes in the RWM. After perforation experiments, confocal images of the RWM show no tearing or ripping and the SEM images of the microneedles indicate no damage on the tips or shafts. The resistive force exerted by the RWM on the microneedle is around 2 mN, which is two orders of magnitude smaller than the catastrophic rupture strength of a glassy carbon micropillar.

The glassy carbon microneedles reported herein will enable clinicians to safely introduce multiple perforations of consistent size across the RWM to enhance and control the rate of delivery into the cochlea of therapeutic agents injected into the middle ear space intratympanically while minimizing leakage of perilymph from the cochlea into the middle ear. As such, it can be an important medical tool to treat a wide variety of hearing and balance disorders.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.