Near-Field Enhanced Plasmonic-Magnetic Bifunctional Nanotubes for Single Cell Bioanalysis

surface-enhanced Raman scattering (SERS). More than 2 × SERS enhancement is observed from the hollow section compared to the solid section of the same nanotube. The substantial SERS enhancement on the hollow section is attributed to the dual-sided coating of Ag NPs as well as the near-ﬁ eld optical coupling of Ag NPs across the nanotube walls. Experimentation and mod-eling are carried out to understand the dependence of SERS enhancement on the NP sizes, junctions, and the near ﬁ eld effects. By tuning the aspect ratio of the embedded nanomagnets, the magnetic anisotropy of nanotubes can be readily controlled to be parallel or vertical to the long directions for nano-manipulation. Leveraging the bifunctionality, a nanotube is magnetically maneuvered to a single living mammalian cell amidst many and its membrane composition is analyzed via SERS spectroscopy.


Introduction
Bifunctional plasmonic-magnetic nanoparticles (PM-NPs) are unique hybrid nanomaterials consisting of both optical and magnetic components in a rationally designed nanoscale architecture and have recently attracted intense research interest. [ 1 ] Possessing both enhanced optical and magnetic properties, PM-NPs can be extremely useful for biomedical applications that require either optical sensing/imaging/heating, magnetic stimulation/manipulation, or both functionalities. [ 1a , b , d , 2 ] For instance, PM-NPs can attach to biological entities such as cells and molecules to separate those entities under external magnetic fi elds and simultaneously detect their chemical nature via optical sensing. [ 3 ] The PM-NPs can also be deployed to study the mechanical properties of deeply embedded biological tissues by magnetic fi eld-induced mechanical stimulation [ 4 ] and monitoring the responses by in-situ optical imaging. [ 5 ] However, current available bifunctional PM nanostructures have been largely limited to quasi-zero-dimensional (0D) nanostructures, such as nanospheres and nanoshells. [ 1a-d , 2a ] To our knowledge, there are few reports on quasi-one-dimensional (1D) PM nanostructures, although 1D PM nanostructures provide exclusive advantages for biomedical applications that are unavailable for 0D nanostructures.
In this report, we present a unique type of 1D PM nanotubes and demonstrate their application in targeted singlecell sensing. The PM nanotubes consist of silica nanotubes with embedded solid Ni nanomagnets and uniformly dual-surface-distributed plasmonic Ag NPs. The PM nanotubes provide a high hotspot density (approximately 1200/ μ m 2 on the outer surface) at the junctions of Ag NPs for SERS biodetection. The solid embedded sections of nanotubes provide single-molecule sensitivity with an enhancement factor up to 7.2 × 10 9 . More than 2 × SERS enhancement was observed from the hollow sections than that from the solid section of the same nanotube. This substantial SERS enhancement is induced by the double sided coating of Ag NPs on the nanotubes as well as the near-fi eld optical coupling between Ag NPs on the inner and outer surfaces of the nanotubes. The dependence of SERS enhancement on the particle sizes, junctions, and the near fi eld effects was carried out by both experimentation and modeling. The magnetic anisotropy of the nanotubes, due to the embedded nanomagnets, can be readily tuned to be parallel or vertical to the long direction of the nanotubes for controlled manipulation. Leveraging the nanotubes' unique bifunctionality, we magnetically maneuvered a nanotube to a living Chinese hamster ovary cell and detected the membrane composition of the specifi c cell with SERS spectroscopy. These bifunctional nanotubes are desirable for multiple-task applications in single-cell bioanalysis, biochemical detection, imagingcontrast enhancement, magnetic manipulation and separation, and biosubstance delivery.
There are three outstanding features provided by the 1D PM nanotubes: First, the unique longitudinal geometry of nanotubes is compatible with both biological cells and biomolecules in terms of length and diameter. For instance, the lengths of magnetic anisotropies (thickness of 30 nm) can be readily embedded in the nanotubes (to be discussed) (Figure 1 d). On the entire outer surface of nanotubes, arrays of Ag NPs were uniformly distributed (Figure 1 e). The Ag NPs also grew on the interior surfaces of the nanotubes as shown in the SEM images obtained by cross-sectional focused ion beam (FIB) milling (Figure 1 f). The Ag NPs were semi-spherical and densely arranged, yet overlapping NPs were rarely found.
The Ag NP sizes and junctions can be optimized for highly sensitive SERS detection. With fi xed volume of nanowire suspension (5.7 × 10 8 /mL, 400 μ L) and PVP (10 mL of 2.5 × 10 − 5 M in ethanol), we systematically varied the total volume of AgNO 3 (0.06 M) and NH 3 · H 2 O (0.12 M) (v:v 1: 1) from 20 μ L to 1200 μ L, i.e. 20 μ L (0.2 × sample), 600 μ L (6 × sample), 800 μ L (8 × sample), and 1200 μ L (12 × sample). The morphologies of the as-synthesized nanotubes showed distinctive particle and junction sizes (Supporting Information Figure S2). The average diameters of the Ag NPs increased from 10.2 ± 2.4 nm (0.2 × sample) to 24.8 ± 6.7 nm (8 × sample) with the volume of the AgNO 3 /NH 3 · H 2 O solution. The density of Ag NPs reached a maximum in the 6 × sample (2050/ μ m 2 ). The detailed characterization is given in Supporting Information nanotubes can be adjusted to tens of micrometers for efficient attachment, manipulation, and separation of cells. [ 6 ] The nanoscale diameters restrict the number of molecules that one nanotube can interact with, which is important for moleculelevel biosensing and drug delivery. [ 7 ] Second, substantially enhanced plasmonic properties are sensitively obtained in the nanotubes. The plasmonic NPs coated on the entire surfaces of the nanotubes provide large and uniform SERS EFs, similar to those frequently reported in patterned substrates, [ 8 ] which are provided by traditional 0D plasmonic NPs or their aggregates. Third, by controlling the aspect ratio of the embedded Ni nanosegment, the magnetic moment and anisotropy can be facilely tuned to the desired value, which is important for effi cient magnetic separation and manipulation. [ 9 ] 2. Results and Discussion

Design and Fabrication
In order to synthesize such PM nanotubes, a rationally designed four-step approach has been used: 1) multi-segment Ag/Ni/ Ag (3/3/3 μ m) nanowires were electrodeposited as growth templates for silica nanotubes ( Scheme 1 a and Figure 1 a); 2) a layer of silica, with controlled thickness of 70 nm, was uniformly plated on the outer surface of the Ag/Ni/Ag nanowires (Scheme 1 b) via hydrolysis of tetraethyl orthosilicate; [ 10 ] 3) the Ag segments were selectively etched, resulting in the hollow silica nanotubes with magnetic Ni embedment (Scheme 1 c and Figure 1 b); 4) plasmonic Ag NPs were uniformly coated through PVP assisted catalysis [ 11 ] on the inner and outer surfaces of the nanotubes with optimized particles and junction sizes (Scheme 1 d and Figure 1 c,e). For detailed fabrication process, see experimental section.
Scanning electron microscopy (SEM) images show that hollow nanotubes with embedded cylindrical solid segments were successfully synthesized (Figure 1 c). Energy dispersive X-ray spectroscopy (EDS) confi rmed that the cylindrical solid was Ni (Supporting Information Figure S1). By using the same method, multiple Ni nanodisks with controlled   increased with the concentration of R6G (Figure 2 a,b from the 8 × sample). As low as 100 fM R6G can be decisively determined with a signal-to-noise ratio of 7.2 (Figure 2 c). The SERS EF values were estimated by following a well-established approach. [ 14 ] The highest EF of 7.2 × 10 9 was obtained from the 8 × sample on the Ni embedded sections (Supporting Information Figure S6). According to bianalyte studies, an EF value of above 10 9 is suffi cient for single-molecule sensing of various biochemical species. [ 3,15 ] The sensitivity of our nanotubes is already on the single molecule levels.
Interestingly, we also noticed that Ag NPs not only grew on the outer surface of the nanotube, but also on the interior surface of the nanotube as shown in the SEM images of crosssections obtained by focused ion beam (FIB) milling (Figure 1 f). Therefore, there were double-layer Ag NPs on the hollow parts of the PM nanotubes and single-layer Ag NPs on the solid part of the nanotubes (Scheme 1 inset). Due to characterization diffi culties, we cannot directly measure the size and distribution of the inner layer Ag NPs. However, the effect of the inner surface coating of Ag NPs, can be known from SERS measurement. When a laser beam scanned along a single nanotube with 70 nm thick silica coating, the hollow segments of nanotubes with double-layer Ag NPs exhibited more than two-time SERS intensity than those from single-layer Ag NPs on the outer surface of the Ni embedded section of nanotubes (Figure 2 g). Similar results were also observed when we replaced the Ni segment with Pt (Figure 2 h). The experimental details are provided in Supporting Information Figure S5.

Plasmonic Simulation
It is interesting to understand how the EF depends on Ag particle and junction sizes, as well as how the Ag NPs on the Figure S3, S4 and S7. The variation of the particle sizes may be attributed to the dynamic competition between nucleation and crystalline growth of Ag NPs, which has been commonly observed in NP growth. [ 12 ]

SERS Characterization
The SERS performance of the nanotubes were characterized and understood by experimentation and numerical simulation. It is known that a laser-beam focused on two closely neighboring Ag NPs can generate high-intensity electric ( E ) fi eld in the narrow junction (a few nm) due to localized surface plasmonic resonance as a result of coherent electron oscillation in the Ag NPs. The junction with enhanced E -fi eld is often referred as "hotspot". If a molecule is in the vicinity of a hotspot, its Raman scattering signals can be signifi cantly amplifi ed with | E | 4 dependence. This phenomenon is the so-called surface enhanced Raman scattering (SERS). [ 13 ] SERS EF is largely determined by E -fi eld intensity and thus the sizes of junctions. If we only consider extremely narrow junctions ( < 2 nm) on the nanotubes, which contribute most to SERS enhancement, the density of hotspots on the outer surface of nanotubes went up from ≈ 0/ μ m 2 for the 0.2 × sample to the maximum of 1200/ μ m 2 for the 8 × sample (Supporting Information Figure S6). The corresponding average hotspot sizes can also be determined. The 8 × samples provided the smallest average size of hotspots (1.16 nm) and the 6 × sample gave the largest (1.4 nm) as shown in Figure 2 d and Supporting Information Figure S6.
To characterize the SERS performance of nanotubes, we employed standard Raman probes including Rhodamine 6G (R6G) (1 μ M to 1 fM) and 1,2-bi-(4-pyridyl) ethylene (BPE) (1 mM). All of the nanotubes can detect R6G and BPE molecules except the 0.2 × sample (Figure 2 a,e). SERS intensity It is known that the SERS enhancement can be attributed to two factors: E -fi eld enhancement due to the plasmonic resonance of NPs and chemical enhancement due to charge transfer between the molecules and metal particles. [ 3 ] The E -fi eld enhancement can be approximated as where ω L is the resonant angular velocity of the local fi eld E in an external fi eld of E 0 . Due to the quadruple dependence on the local E fi eld, the E -fi eld enhancement is usually considered to be a major contributor to SERS. We used a 3D fi nite difference time domain (FDTD) method to simulate this effect. We simulated the normalized electric fi eld ( | E | / | E 0 | ) close to the Ag NPs. The excitation wavelength was set at 532 nm with inplane polarization perpendicular to the nanotube surface. To fi nd the effect of NP junctions on SERS EF, we began with the simulation for a single particle, a dimer, and then a 4 × 4 array ( Figure 3 a-c, the shape of Ag NPs was assumed to be cylindrical with a uniform thickness of 20 nm. The junction was set at 3 nm). A single silver NP provided electric fi eld enhancement of | E | / | E 0 | = 2.3, hence the SERS EF = 28 due to the E 4 dependence; metal dimers and a 4 × 4 array provided | E | / | E 0 | = 8.6 (EF SERS = 5470) and | E | / | E 0 | = 9.1 (EF SERS = 6857), respectively. This simulation confi rms that a group of metallic particles on the same surface can provide stronger E fi eld enhancement than that of single particles or dimers.
A group of NPs with controlled distributions should further enhance the E fi eld. Therefore we simulated the E -fi eld enhancement depending on junctions of the particle arrays. As plotted in Figure 3 d, narrower junctions between NPs exhibit a strong electric fi eld for an array with Ag NPs of 25 nm in diameter. At a fi xed junction of 1 nm, the highest electric fi eld enhancement is obtained in particle arrays with diameters between 20-50 nm as shown in Figure 3 e. The simulation results excellently agree with previous work, which refers to such results as proof of the "extrinsic size effect". [ 16 ] For NP with larger sizes ( > 40 nm), the E -fi eld enhancement reduces because the plasmonic resonance shifts to longer wavelength. When such NPs are excited by short-wavelength lasers, they induce higher-order electron cloud distortion of conduction electrons and thus degrade the plasmonic resonance. For NPs that are less than 25 nm, scattering of electrons from the particle's surfaces produces a damping term that is inversely proportional to the particle diameter. This means that more optical energy is converted to heat instead of being scattered to produce local fi eld enhancement. As a result, by comparing the electric fi elds among NPs with different diameters and junctions, we have optimized the design of the Ag NPs. The result of simulation readily explained our experimental observation: among all the samples, the 8 × samples with the largest diameters (24.8 ± 6.7 nm) and smallest gaps ( ≈ 1.16 nm) offered the highest SERS EF. To understand how the dual-side-Ag-coated hollow nanotubes can further enhance SERS than the single-side-Ag-coated nanotubes with solid embedment do, we carried out numerical simulations by Comsol 3.5a RF module. In our modeling, a 3D silica nanotube is constructed (illustrated in Scheme 1 d and insets): the inner cylinder radius is 150 nm, and the shell thickness is 70 nm. The densely coated Ag NPs are simplifi ed by a 2D conformal array attached to the outer and inner surfaces of the silica nanotubes. The Ag NP diameter is 25 nm and the gap between them is 2 nm. The silica nanotube is placed on top of a glass substrate, and is excited by a surface normal Gaussian beam with beam diameter of 1 μ m at 532 nm wavelength. The polarization direction is perpendicular to the axis of the cylinder. Table 1 lists our simulation results corresponding to our experimentally measured devices and Figure 4 shows the cross sectional views of the electric fi eld distribution of the four devices as listed in Table 1 .
In Figure 4 a, we surprisingly fi nd that the hot-spot with maximum electric fi eld enhancement is actually at the bottom of the nanotube, not at the top. This is because the bottom hot-spot is surrounded by high-index silica; while the other hot-spots have only one side contacted with silica (other side is exposed to low-index air). Additionally, the inter-particle coupling through the NP chain at the outer surface enhances the electric fi eld at the bottom. [ 17 ] Comparing Figure 4 a,b, we are able to clearly conclude the contribution of inner layer Ag NPs: the presence of inner layer not only adds more hot spots for SERS sensing, but also signifi cantly increase the intensity of  Moreover, we noted that the thickness of silica also affects the enhancement of SERS. We observed near-fi eld enhancement effect on nanotubes with silica coating ranging from 70 to 150 nm. However, when the thickness of silica was increased to 300 nm, the near-fi eld enhancement effect was not observed, which may be attributed to the reduced plasmonic coupling between Ag NPs across the silica shell.

Magnetic Characterization
Not only plasmonically sensitive, the unique nanotubes also offered tunable magnetic properties for controlled manipulation. The magnetic anisotropy of Ni segment is dominated by its shape anisotropy [ 9 ] as opposed to its weak crystalline anisotropy. When the aspect ratio of Ni segment is high, e.g. 10/1 (Ni length 3 um, diameter 300 nm), the anisotropy direction and the easy axis is along the nanotube long axis as measured by vibrating sample magnetometry (VSM) ( Figure 5 a). A hysteresis loop along magnetic easy axis demonstrates higher magnetic remanence and squareness than those measured perpendicular to the nanotubes. When the aspect ratio of the Ni segments is below 1, e.g., a stack of thin Ni disks with diameters of 300 nm and thickness of 30 nm, magnetic anisotropy is generally transverse to the nanotubes with essentially zero remanence due to the anti-parallel coupling of the magnetizations in neighboring nanodisks (Figure 5 b). This fascinating way of controlling the magnetic anisotropy has been vividly demonstrated by manipulating nanotubes in suspension with a magnetic fi eld. As shown in Figure 5 , nanotubes with magnetic anisotropy along the long axis align with the magnetic fi eld during transport, but those with transverse magnetic anisotropy align perpendicular to the magnetic fi eld. The transport speed was 2-7 μ m/s, which can be controlled by the magnetic fi eld gradient.

Single-Cell Bioanalysis
Although single complex biological samples can be investigated with standard Raman microscopy, a detailed investigation of specifi c components on the cell surface is not possible with this approach. [ 19 ] In our work, we demonstrated the utility of the bifunctional nanotubes in revealing the the hot spots in the outer layer NPs, which is due to the nearfi eld effect [ 18 ] of inter-layer coupling between NP chains in the inner and outer surfaces of the nanotubes. When the air core is fi lled with Pt, which has no surface plasmonic resonance at visible wavelength, we observed interesting phenomena: fi rst, the EF of the hot-spots on top of the nanotube is enhanced roughly 2 × , which is due to the refl ected light from the Pt core; second, the hot-spots at the bottom of the nanotube, however, is signifi cantly reduced because almost no light can penetrate the Pt core to excite the surface plasmons. And moreover, the interparticle coupling through the NP chain can be weakened by evanescent fi eld absorption of Pt. Similar electric fi eld enhancement is observed in Ni-fi lled nanotube as well.
If we assume there are enough molecules so that every hotspot can contribute to SERS measurement, we can calculate the total SERS signals by ∑ | E i | 4 , where E i is the electric fi eld in each hot spot. The total SERS signals of these four devices are 0.8 × 10 5 , 2.5 × 10 6 , 1.4 × 10 5 and 1.3 × 10 5 a.u., respectively. It is seen that device 2 will be able to provide more than 10 × higher SERS signals than device 3 and 4. However, in reality, the SERS signals from hot-spots at the bottom to the nanotubes are more diffi cult to collect due to NP scattering. Experimental we obtained more than 2 × differences (Figure 2 g,h). The comparison of single-side and dual-side Ag NP coated segments in the nanotubes is summarized in Supporting Information Figure S8.   membrane composition of a single Chinese hamster ovary (CHO) cell amidst many. We chose CHO cells because they are widely used in biological research, especially in studies of genetics, toxicity screening, gene expression, and expression of recombinan proteins. Here, leveraging the unique bifunctionality of the nanotubes, we can precisely transport a nanotube to a specifi c living CHO cell (see Experimental Section) amidst many and detected its membrane chemistry with SERS spectroscopy. A PM nanotube was transported and aligned in the direction of the magnetic fi eld and precisely landed on the membrane of a CHO cell (overlapped images in Figure 6 a,b and Supporting Information Video S9). From SERS spectra (Figure 6 c), which was taken from the nanotube with an integration time of 5 s, shows strong characteristic peaks of lipids. [ 7 , 19 , 20 ] The peak position 1511 cm − 1 can be assigned to amide II, which is from protein (blue bar in Figure 6 c). This results revealed that the cell membrane in contact with the nanotube consists mostly lipids and some protein molecules, which is consistent with real cell membrane composition. [ 19 ] Without nanotubes, no Raman signals can be detected from the cell, this clearly demonstrating the highly desirable bifunctionality of the nanotubes for precision and ultrasensitive single-cell bio-analysis. This technique is generally applicable to any adhesive live cells. It can be readily applied to hamster cells as well as to mouse or human cells.

Conclusions
In summary, we have successfully designed and synthesized a unique type of near-fi eld enhanced bifunctional PM-active nanostructure that consists of a hosting silica nanotube, a magnetic segment embedded within the nanotube, and Ag NPs uniformly coated on the dual surfaces of the nanotube. By controlling the fabrication conditions, both the diameter and junction of Ag NPs can be precisely controlled for ultrasensitive molecular sensing. The 3D FDTD simulation of E -fi eld enhancement agrees with the experimental results. Higher SERS intensity is found on hollow than the solid parts of the PM nanotubes, and it is confi rmed to be from the near fi eld coupling between the inner and outer layer of Ag NPs. The embedded nanomagnets with tunable magnetic anisotropy allow fl exible manipulation of the nanotubes with external magnetic fi elds. Such bifunctional nanostructures can be transported to a living Chinese hamster ovary cell amidst many other cells to reveal the membrane composition. The PM nanotubes are suitable for single-cell bioanalysis as well as various biological applications, such as biochemical sensing, magnetic manipulation, separation, MRI contrast, and biosubstance delivery.

Experimental Section
Fabrication of Plasmonic-Magnetic Nanotubes : A series of strategies was implemented to synthesize the unique bifunctional nanotubes using the above approach. In brief, the multisegment Ag/Ni/Ag nanowires were fabricated by electrodeposition in nanoporous anodized aluminum oxide (AAO) templates as reported elsewhere. [ 11,21 ] In brief, a Cu layer of about 500 nm in thickness was thermal evaporated onto the back of the template to seal the pores and serve as the working electrode in a three-electrode electrodeposition system. The electrodeposition of metal materials gradually fi ll the bottom of the nanopores working electrode to form nanowires. Finally, the AAO template was dissolved in 2 M NaOH solution to release the free-standing nanowires. The amount of electric charge passing through the circuit controls the length of nanowires to 7 nm. [ 22 ] The pore size of nanoporous template controls the diameters of the nanowires from 20 to 400 nm, with different compositions along the lengths (e.g., Ag/Ni/Ag nanowires) (Figure 1 a), and here we prepared 300 nm diameter 3/3/3 μ m Ag/Ni/Ag nanowires ( − 1.1 V for Ag and − 1 V for Ni).
Next, the Ag/Ni/Ag nanowires were used as templates for fabricating Ni-embedded silica nanotubes. We fi rst coated an amorphous silica layer on the surface of the Ag/Ni/Ag nanowires. Silica was used due to its porous structure with a high surface area, biocompatible properties, and drug carrier capability. [ 10a ] Here, it also serves as a support substrate for plasmonic Ag NPs. The reaction was accomplished by hydrolysis of tetraethyl orthosilicate for 2-5 h with a controlled thickness of a few to hundreds of nanometers. [ 10 ] Next, the Ag segments were selectively etched with a mixture (4:1:1) of methanol (99%), hydrogen peroxide (30%), and ammonia hydroxide (28-30% as NH 3 ), which result in Ni-embedded nanotubes as shown in Figure 1 b.
Finally, arrays of plasmonic Ag NPs were uniformly synthesized on the surface of silica nanotubes by reduction of Ag ions with PVP (10 mL, 2.5 × 10 − 5 M in ethanol) from a mixed solution of silver nitrate (0.06 M, 400 μ L) and ammonia hydroxide (0.12 M, 400 μ L) at 70 ° C for 7 h.
Cell Culture and Reagents : CHO cells (ATCC) were cultured in RPMI medium (Invitrogen) supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin (Invitrogen). Cells were maintained in a humidifi ed 37 ° C, 5% CO 2 incubator. Before the experiment being conducted, CHO cells were washed by phosphate buffered saline (PBS) followed by addition of nanotubes (dispersed in PBS).

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. Figure 6 . a,b) Transport a nanotube to a single CHO cell amidst many. c) SERS spectrum from the CHO cell membrane is dominated by lipid contribution, and a few peaks can be assigned to protein (blue bar).