Micro.uidic devices for terahertz spectroscopy of biomolecules 
Paul A. George1, Wallace Hui1, Farhan Rana1, Benjamin G. Hawkins2 , 
A. Ezekiel Smith3, Brian J. Kirby4 
1School of Electrical and Computer Engineering, Cornell University, Ithaca, NY, 14853
2Department of Biomedical Engineering, Cornell University, Ithaca, NY, 14853
3School of Applied and Engineering Physics, Cornell University, Ithaca, NY, 14853
4School of Mechanical and Aerospace Engineering, Cornell University, Ithaca, NY, 14853

pag25@cornell.edu 
Abstract: We demonstrate micro.uidic devices for terahertz spectroscopy of biomolecules in aqueous solutions. The devices are fabricated out of a plastic material that is both mechanically rigid and optically transparent with near-zero dispersion in the terahertz frequency range. Using a low-power terahertz time-domain spectrometer, we experimentally measure the absorption spectra of the vibrational modes of bovine serum albumin from 0.5 -2.5 THz and .nd good agreement with previously reported data obtained using large-volume solutions and a high-power free-electron laser. Our results demonstrate the feasibility of performing high sensitivity terahertz spectroscopy of biomolecules in aqueous solutions with detectable molecular quantities as small as 10 picomoles using micro.uidic devices. 
© 2008 Optical Society of America 
OCIS codes: (300.0300) Spectroscopy; (280.0280) Remote sensing and sensors; (230.0230) Optical devices; (17.0170) Medical optics and biotechnology 
References and links 
1.	
S. Hayward and N. Go, “Collective variable description of native protein dynamics,” Annu. Rev. Phys. Chem. 46, 223–250 (1995). 

2.	
M.C. Chen and R.C. Lord, “Laser-excited Raman spectroscopy of biomolecules: conformational study of bovine serum albumin,” J. Am. Chem. Soc. 98, 990–992 (1976). 

3.	
W. Zhuang, Y. Feng, and E.W. Prohofsky, “Self-consistent calculation of localized DNA vibrational properties at a double-helix–single-strand junction with anharmonic potential,” Phys. Rev. A 41, 7033–7042 (1990). 

4.	
A.G. Markelz, and A. Roitberg, and E.J. Heilweil, “Pulsed terahertz spectroscopy of DNA, bovine serum albumin and collagen between 0.1 and 2.0 THz,” Chem. Phys. Lett. 320, 42–48 (2000). 

5.	
T.R. Globus, D.L. Woolard, T. Khromova, T.W. Crowe, M. Bykhovskaia, B.L. Gelmont, J. Hesler, and A.C. Samuels, “THz-spectroscopy of biological molecules,” J. Bio. Phys. 29, 89–100 (2003). 

6.	
T.R. Globus, D.L. Woolard, T.W. Crowe, T. Khromova, M. Bykhovskaia, B.L. Gelmont, and J. Hesler, “Terahertz Fourier transform characterization of biological materials in a liquid phase,” J. Phys. D 39, 3405–3413 (2006). 

7.	
J. Xu, K.W. Plaxco, and S.J. Allen, “Probing the collective vibrational dynamics of a protein in liquid water by terahertz absorption spectroscopy,” Prot. Sci. 15, 1175–1181 (2006). 

8.	
M. Nagai, H. Yada, T. Arikawa, and K. Tanaka, “Terahertz time-domain attenuated total re.ection spectroscopy in water and biological solutions,” Intl. J. Inf. and Mill. Waves 27, 505–515 (2006). 

9.	
J. Kitagawa, T. Ohkubo, M. Onuma, and Y. Kadoya, “THz spectroscopic characterization of biomolecule/water systems by compact sensor chips,” Appl. Phys. Lett. 89, 041114 (2006). 

10.	
T. Baras, T. Kleine-Ostmann, and M. Koch, “On-chip THz detection of biomaterials: a numerical study,” J. Bio. Phys. 29, 187–194 (2003). 

11.	
M. Nagel, P.H. Bolivar, M. Brucherseifer, H. Kurz, A. Bosserhoff, and R. B¨uttner, “Integrated THz technology for label-free genetic diagnostics,” Appl. Phys. Lett. 80, 154 (2002). 

12.	
“Zeonor Production Information Sheet,” Zeon Corporation (2004). 

13.	
T.I. Wallow, A.M. Morales, B.A. Simmons, M.C. Hunter, K.L. Krafcik, L.A. Domeier, S.M. Sickafoose, K.D. Patel, and A. Gardea, “Low-distortion, high-strength bonding of thermoplastic micro.uidic devices employing case-II diffusion-mediated permeant activation,” Lab on a Chip 7, 1825–1831 (2007). 

14.	
B.G. Hawkins, A.E. Smith, Y.A. Syed, and B.J. Kirby, “Continuous-.ow particle separation by 3D insulative dielectrophoresis using coherently shaped, dc-biased, ac electric .elds,” Anal. Chem. 79, 7291-7300(2007). 

15.	
“Bovine Serum Albumin Product Information Sheet,” Sigma-Aldrich (2000). 

16.	
R.D. Levine, Molecular Reaction Dynamics, Cambridge Univeristy Press (2005). 


1. Introduction 
Important biomolecules, including proteins, RNA, and DNA have vibrational modes with frequencies in the 0.1 THz -5 THz range [1, 2, 3, 4, 5, 6, 7, 9, 10, 11]. These modes correspond to collective molecular oscillations and rotations as well as the relative motion of molecular subdomains [1, 2, 3]. 
Several studies of biomolecules in the THz frequency range have been conducted [4, 5, 6, 7, 8, 9, 10, 11]. Due to strong water absorption at THz frequencies, these studies have typically been limited to dry or partially hydrated specimens. In [4], THz time-domain spectroscopy (THz-TDS) was performed on dried .lms of biomolecules that were up to 7.5 mm thick. In [5, 6], THz-TDS of hydrated liquid-phase biomolecules prepared in gel .lms was performed. Studies of the THz absorption of biomolecules in large volume aqueous solutions were conducted in [7], but these experiments required the use of a high-power free electron laser. Unfortunately, such high-power THz sources are not readily available. THz-TDS of smaller volumes of sucrose solutions was demonstrated using attenuated total re.ection spectroscopy [8]. Recently, integrated spectroscopy systems based upon microstrip waveguides have been proposed and demonstrated in [9, 10, 11]. The performance of these devices suffers from the limited bandwidth of microstrip components. 
In this paper, we demonstrate micro.uidic channels for THz spectroscopy of biomolecules. Performing spectroscopy of biomolecules in micro.uidic devices can be advantageous for the following reasons: (1) biological specimens can be interchanged or reacted in real-time; (2) the dimensions of micro.uidic channels can be designed to facilitate the use of low-power THz-TDS systems by avoiding excess water absorption and to enable the spectroscopy of picomole quantities of biomolecules; and, (3) micro.uidic channels can easily be integrated with photonic components to realize multifunctional spectroscopy platforms. A schematic of a THz sensing platform based on micro.uidic channels is shown in Fig. 1. The realization of such devices requires materials that have low loss and dispersion from optical to THz frequencies, are mechanically stable and chemically inert, and are compatible with microfabrication. 
Using micro.uidic devices, we measure the absorption coef.cient of the protein bovine serum albumin (BSA) in the 0.5 THz -2.5 THz frequency range with a detection sensitivity approaching 1.5 micromole/mL. The measured spectrum is in excellent quantitative agreement with the previously reported results in [7]. Our results demonstrate the feasibility of performing THz spectroscopy of picomole quantities of biomolecules in micro.uidic devices using sub-microWatt low-power THz sources. 
2. Micro.uidic devices for THz spectroscopy 
Micro.uidic channels are commonly fabricated from polydimethylsiloxane elastomer (PDMS), which has high water content and therefore exhibits large absorption at THz frequencies. The micro.uidic devices used in this work were fabricated from Zeonor 1020R (Zeon Corporation), a cyclo-ole.n polymer that has high transmission from UV to Far-IR wavelengths, is mechanically robust, and has signi.cantly lower water content than PDMS [12]. 
Terahertz time-domain scans of a 2 mm slab of Zeonor 1020R and a 1 mm slab of PDMS 
THz pulse 

Fig. 1. A system for on-chip THz sensing. The device contains sample and reagent inputs, a reaction chamber, and a detection chamber for THz as well as optical and IR spectroscopy. 
were conducted using a THz-TDS based on a <100> n-InAs (5×1016 cm-3) emitter and a 1mm <110> ZnTe electro-optic detector. The emitter was pumped by 1 W of optical power from a 90 fs Ti:Sapphire ultrafast laser (780 nm) with an 86 MHz repetition rate. The THz pulses had a bandwidth of 0.25 THz -3 THz and the THz-TDS had a power SNR approaching 
106. To eliminate absorption from water vapor, the entire experimental setup was encased in a nitrogen-purged plastic box. The measured error in the repeatability of successive scans was less than 5% and is attributable to the drift of optomechanical components and the laser beam. 
Because both the Zeonor and PDMS slabs were thick and exhibited only broad spectral features, we limited the spectrometer scan length so as to avoid etalon effects caused by multiple internal re.ections. The index of refraction and absorption coef.cient of Zeonor and PDMS were extracted using Fresnel’s equation for the frequency-dependent amplitude transmission through a dielectric slab, t, without multiple internal re.ections 
t = EEmr((..))=(1 +4nn)2 ei .c (n-1)d (1) 
In Eq. 1, Em(.)is the measured THz spectrum of the pulse after propagation through the slab, Er(.)is the THz spectrum of the reference pulse, d is the slab thickness, and n is the complex index of refraction of the slab. The absorption coef.cients of Zeonor and PDMS calculated from the extracted complex indices of refraction are shown in Fig. 2. 
Zeonor 1020R has a measured index of approximately 1.518 and an absorption coef.cient <1cm-1 from 0.5 THz -2.5 THz. The low absorption of Zeonor over a large frequency range makes it better suited for terahertz spectroscopy than PDMS. 
Fabrication of micro.uidic devices out of Zeonor 1020R has previously been described in [13, 14] and is shown pictorially in Fig. 3(a) and (b). The Si template for embossing the .uidic channel in Zeonor was constructed using standard photolithography techniques and deep Si etching. The .nal device was 5 mm wide, 25 mm long, and 95 µm deep. A 14.5 cm 2 square Zeonor piece was embossed with the template on a hotpress at a temperature of 115 C and a pressure of 2.55 MPa. The piece was then cooled under load until its temperature was below the glass transition temperature (102 C [12]). Fluid input and output ports were then de.ned. Next, the embossed piece and a blank 14.5 cm2 Zeonor square were soaked in an ethanol/decalin (80/20) solution for 30 s and 1 min, respectively. The pieces were then rinsed in ethanol, dried and bonded on a hotpress at a temperature of 65 C and a pressure of 1.6 MPa for 40 min to form the channel. A picture of a typical completed device is shown in Fig. 3(b). 
0.5	1.0 1.5 2.0 2.5 Frequency (THz) 
Fig. 2. The measured absorption coef.cient of Zeonor 1020R and PDMS. Zeonor has a nearly constant index of 1.518 (not shown) and an absorption coef.cient <1cm-1 at THz frequencies, which is 10-20 times smaller than that of PDMS. 
T = 115 C, P = 2.55 MPa 


(a)	(b) 
Fig. 3. Fabrication of the micro.uidic devices used in this work. (a) A slab of Zeonor 1020R was .rst embossed using a Si template. (b) The embossed slab was then bonded to another piece of Zeonor 1020R. The .nal channel depth was 95 µm. 

3. Measurement of the THz absoprtion spectrum of BSA in micro.uidic channels 
As a proof of principle, we measured the absorption spectrum of the protein bovine serum albumin (BSA) in Zeonor micro.uidic channels. BSA is a well studied proteins and has previously been used in the study of biomolecules at THz frequencies [4, 7, 9]. It is a polypeptide chain consisting of 583 amino acids [15] with a sequence homology similar to Human Serum Albumin [7]. It has an ellipsoidal geometry [2] with a 54% a-helix and 18% ß-sheet structure and has a molecular weight of approximately 66.430 kDa [15]. 
Aqueous solutions of globulin-free and fatty-acid-free lyophilized BSA powder (Sigma-Aldrich A0281 Batch #075K7545) were prepared in a 0.05 M phosphate buffer (pH = 7.5) at concentrations of 101 mg/mL (pH = 7.0), 200 mg/mL (pH = 6.9), and 305 mg/mL (pH = 6.9) as measured by 279 nm UV spectrophotometry. At these high concentrations, the uncertainty in the measurement of the BSA concentration was approximately ±10%. 
The complex THz transmission spectra of the empty micro.uidic channel, the channel .lled with phosphate buffer, and the same channel .lled with the BSA solutions were obtained in succession. To avoid perturbing the experimental setup, the solutions were pumped into the channel with syringes connected to long Tygon tubing. Scans of the phosphate buffer with a 
0.02 THz resolution and BSA solutions exhibited no signi.cant features with spectral width less than 0.15 THz between 0.5 THz -2.5 THz. This observation is consistent with the THz absorption spectra of water and BSA reported in [4, 7, 9]. 
The THz transmission of each solution was averaged over three successive scans with a measured reproducibility near 98%. Equation 1 is not valid for extracting the absorption coef.cient from the measured THz transmission because multiple internal re.ections inside the 95 µm channel cannot be ignored. Instead, we have used Eq. 2 which includes these effects but neglects multiple re.ections from the thick Zeonor slabs on either side of the channel. 

Es(.) i .(ns-1)dc (nz +1)2 -(nz -1)2e2i .c dc 
t == nsec	(2)
Ea(.)	(nz +ns)2 -(nz -ns)2e2i .c nsdc 
Es(.)and Ea(.)are the measured THz spectra of the pulse after propagation through the channel .lled with solution and with air, respectively. dc is the channel depth, nz and ns are the complex indices of Zeonor and the solution, respectively. 
The extracted absorption coef.cients of the phosphate buffer and BSA solutions are shown in Fig. 4. Consistent with the experimental results in [9], a monotonic decrease in the absorption coef.cient of the BSA solution with increasing concentration is observed. This is due to the expulsion of solvent molecules by less-absorbing BSA molecules. The change in absorption coef.cient between the 0.05 M phosphate buffer and the 101 mg/mL BSA solution is 15 cm -1 -20 cm-1,or8%. 
0.5	1.0 1.5 2.0 2.5 Frequency (THz) 
Fig. 4. The absorption coef.cient of the phosphate buffer and BSA solutions measured by THz-TDS using micro.uidic devices. Values are extracted using Eq. 2. 
The total THz absorption coef.cient of the BSA solution, as, can be written as, 
as = absa Vbsa +apb Vpb (3)Vbsa +Vpb Vbsa +Vpb 
where absa and apb are the absorption coef.cients of the hydrated BSA molecules and the phosphate buffer (Fig. 4), respectively. Vbsa and Vpb are the volumes of the BSA and phosphate buffer in the solution. Based upon density measurements of BSA solutions reported in [7], a molecular radius of 2.8 nm for hydrated BSA was used to calculate Vbsa. Figure 5(a) displays the extracted values of absa as a function of frequency. The error bars indicate the observed ±10% error in the measurement of the concentrations of the BSA solutions. Oscillations in the data for the 101 mg/mL solution near 2 THz are attributable to low SNR and are not present at higher concentrations. The absorption coef.cient plotted in Fig. 5(a) approaches zero at low frequencies and increases nearly monotonically from 0.5 THz -2.5 THz due to the increase in the density of molecular vibrational modes [7]. The values of the absorption coef.cient extracted from measurements of the 101 mg/mL and 200 mg/mL BSA solutions do not exhibit signi.cant concentration dependence in agreement with Beer’s Law [16]. The slight reduction in the extracted absorption coef.cient for the 305 mg/mL solution is most likely due to uncertainties in the measured solution concentration. 
The molar extinction of BSA molecules in aqueous solution (cm -1M-1), ambsa, is related to absa by 
ambsa = absaNAVmbsa	(4) 
where NA is Avogadro’s Number and Vmbsa is the volume of a hydrated BSA molecule in liters [7]. In Fig. 5(b), we compare our measurement of the molecular extinction of BSA to those reported in [7] using large-volume aqueous solutions of BSA and a high-power THz free-electron laser. The good agreement from 0.5 -2.5 THz demonstrates the feasibility of performing THz spectroscopy of biomolecules in aqueous solutions using micro.uidic channels using low-power THz sources. It should be noted no scaling was used to achieve the .t Fig. 5(b) and that the data is absolute, within the indicated error margins. 

0.5	1.0 1.5 2.0 2.5 0.5 1.0 1.5 2.0 2.5 Frequency (THz) Frequency (THz) 
Fig. 5. (a) The measured molecular absorption coef.cient of hydrated BSA molecules. In agreement with Beer’s Law, the absorption coef.cient does not depend on solution concentration. (b) The molar extinction of BSA measured using micro.uidic channels compared to the results in [7]. The excellent agreement demonstrates the feasibility of performing THz spectroscopy of biomolecules in micro.uidic channels using low-power THz sources. 

4. Conclusion 
We have demonstrated for the .rst time terahertz time-domain spectroscopy of bovine serum albumin in aqueous solution using Zeonor micro.uidic channels. The measured results agree well with those previously reported using high-power THZ free-electron lasers [7]. Assuming a minimum BSA solution concentration of 100 mg/mL, a micro.uidic channel of area 500×500 µm2, and a channel depth of 50 µm, the minimum measurable quantity of BSA at 1 THz is approximately 10 picomoles (0.6 µg). Micro.uidic channels offer a new platform for performing THz spectroscopy of small quantities of biomolecules using low-power THz sources. 


Acknowledgments 
The authors would like to acknowledge support from NSF and ARO, as well as helpful discussions with D. Erickson and J. Xu. BJK would also like to thank B. Simmons for sharing protocols from Ref. [13] prior to its publications.