Supporting Information

Experimental details

1. Preparation of Langmuir-Blodgett films

Au (111) substrates were prepared by the Clavilier method[1] and were annealed in a
hydrogen-oxygen flame before every experiment. Au substrates were then placed inside a
UV-ozone cleaner (Ossila, UK) for 10 minutes to remove organic contamination from the
surface. A Langmuir-Blodgett (LB) trough (KSV, Minitrough, Finland) placed on an anti-
vibration table in a dust-reduced environment, was thoroughly cleaned using 5% detergent
followed by copious amounts of water. The Au substrates were then transferred to the LB
trough with the Au surface immersed in the water, which was used as subphase. 30 µl of 1
mg/ml DPPC solution (>99%, Avanti Polar Lipids, USA) in chloroform (99.9%, Acros Organics,
USA) were carefully transferred to the water surface inside the LB trough using a glass syringe.
The chloroform was allowed to evaporate by waiting for 30 minutes. Compression rate was
set to 5 mm/minute and surface pressure was measured using a Wilhelmy balance with a
precision of 0.01 mN/m. Finally, DPPC monolayers were transferred onto Au (111) substrates
at 20 mN/m surface pressure and a transfer speed of 2 mm/minute via the vertical LB
technique. All monolayer transfers were performed at room temperature. At 20 mN/m
surface pressure, DPPC is predominantly present in the liquid condensed (LC) phase.[2]

2. TERS system
TERS measurements were performed using a combined STM/Raman microscope (NT-MDT,
Russia) that combines an STM with a Raman spectrometer and EMCCD detector (Andor,
Ireland).[3] A 632.8 nm He–Ne laser (Spectra-Physics, Newport, Germany) with an output
power of 17 mW (3 mW on the sample) was used as the excitation source. A 100×, 0.7
numerical aperture objective lens (Mitutoyo, Japan) was used for excitation as well as
collection of TERS signals. TERS spectra were measured with a spectrometer grating with 600
lines/mm. AFM topography measurements were performed using peak force tapping mode
AFM (Catalyst; Bruker, USA) with Scanasyst-Air silicon nitride cantilevers (Bruker, USA).

3. Preparation of TERS probes

Electrochemically etched Ag wires (0.25 mm diameter, 99.99% purity, Aldrich, Germany) were
used as STM-TERS probes. A mixture of 1:4 (v/v) perchloric acid: methanol was used as the

1
etchant. Ag wire was first cut to an appropriate length and then immersed into the etching
solution with a 1 mm diameter Pt wire loop acting as counter electrode. A voltage of 10 V was
applied between the electrodes and switched off using a custom-made circuit. Etching was
stopped when a steep drop in current occurred because of the formation of the tip and the
loss of the immersed part of the wire. After etching, the tips were rinsed with methanol and
water and visually inspected using an optical microscope (Nikon, Japan). The tip etching
procedure reproducibly yielded sharp Ag tips with <100 nm diameter that provided a high
TERS signal enhancement and good STM imaging. All TERS experiments were performed using
freshly etched Ag tips.

4. Mass spectrometry measurements

Electrospray ionization (ESI) mass spectra were measured from a freshly prepared DPPC
monolayer on a Au substrate. Control experiments were performed on Au substrate covered
with pure ethanol (Sigma-Aldrich, USA) but without any DPPC. For analysis, the Au surface
containing DPPC was rinsed with 50 µl of pure ethanol and recovered using a pipette. The
collected ethanol was evaporated under nitrogen to obtain a volume of approximately 10 µl
and filled into nanospray emitters for ESI. Nanospray emitters were prepared in-house using
a capillary puller (Sutter Instrument, USA). ESI MS analysis was performed using a Synapt G2S
mass spectrometer (Waters, USA).

5. Data analysis
Data analysis was performed using a custom python script (Python 3.7.5) and Origin software
(OriginLab Corporation, USA). Background subtraction of the raw TERS spectra was carried
out using the asymmetric least squares method. A Satvizky-Goley smoothening was applied
to the spectra measured in the TERS images and the smoothened spectra were deconvoluted
using Gaussian peak fitting with SciPy Python library. The intensity of the fitted Gaussian peak
was used for the calculation of I2930/I2850 ratio. For pixels with I2930 or I2850 signal of less than
20 counts, the I2930/I2850 ratio was set to zero. Deconvolution of averaged TERS spectra was
performed using Origin software. The AFM images were processed using Gwyddion software
(Gwyddion 2.54).

2
Supplementary Figures

Figure S1. Transfer isotherm of the DPPC monolayer transferred at 20 mN/m on Au (111)
surface using the LB technique. The DPPC molecules occupy 0.5 nm2 area per molecule at 20
mN/m surface pressure. Therefore, approximately 3200 and 800 molecules are measured at
each pixel of the TERS images with 40 nm and 20 nm step-size, respectively. This
demonstrates that TERS has the required sensitivity for hyperspectral visualization of a few
hundred DPPC molecules in ambient conditions.

3
Figure S2. (a) AFM topography image of a DPPC monolayer on a Au (111) surface. (b)
Histogram of the AFM topography image in (a) showing a bimodal height distribution. The
height difference between the low (substrate) and high (DPPC) regions is 2.8 nm (in
comparison, the height difference between LC and LE phases is ̴ 1 nm),[4] which confirms the
presence of a predominantly LC phase in the monolayer DPPC film. There are a number of
holes observed in the AFM topography image of DPPC monolayers transferred onto Au (111)
single crystal. Such holes have been observed before by various research groups.[2, 5] They are
a combination of the (1) defects introduced during the transfer of LB film to the substrate[5a],
(2) inhomogeneity of the monolayer at the air-water interface prior to deposition,[6] (3) non-
crystalline residues.[2, 7] Furthermore, the height distribution from the AFM image of DPPC
monolayer transferred to Au (111) single crystal substrate shows a bimodal distribution with
a height difference of 2̴ .8 nm, which is the height of DPPC molecule. This shows that the
predominant phase is LC phase.

4
Figure S3. TERS (red) and far-field Raman (black) spectra of the DPPC monolayer on Au
measured with the TERS probe engaged and retracted from the sample, respectively.
Integration time: 60 s. In the TERS (and SERS) spectrum of DPPC, the intensity of the 2930 cm-
1 band (Fermi resonance associated with the terminal methyl symmetric stretch and the C–H

bending mode) is higher than that of the 2850 cm-1 band (methylene symmetric stretch),
compared to the far-field Raman spectrum.[8] This is due to a modified electronic state of
DPPC via electron transfer from the metal during plasmonic enhancement. Due to the
modification of the electronic state, coupling between the two vibrational modes is boosted,
which enhances the strength of the Fermi resonance.[8]

5
Figure S4. (a)-(c) TERS images of the C-H stretching (2650-3000 cm-1) intensity measured
consecutively in a DPPC monolayer region on Au, to check the stability of the TERS set-up and
the reproducibility of TERS imaging. Integration time: 1 s. Step size: 50 nm. The sample
feature in the bottom right corner is visible in all three TERS images. (b)-(f) Average TERS
spectra measured in the TERS images shown in (a)-(c). Average spectra show a very similar
shape and signal-to-noise ratio of the TERS signal.

6
 Table S1. Assignment of bands in the DPPC Raman spectrum.

Band position (cm-1) Assignment

2850 Methylene symmetric stretch (d+)[9]
2880 Methylene anti-symmetric stretch (d-)[9]
2930 Fermi resonance associated with terminal methyl symmetric
stretch (r⁺FR) [9]
2965 Terminal methyl anti-symmetric stretch (r-)[9b-d]
3040 Anti-symmetric methyl stretch of the choline headgroup[9c, 9d]

7
Figure S5. (a) TERS image of the I2930/I2850 ratio of a patch of DPPC monolayer transferred onto
Au (111) surface as shown in Figure 2f. (b-d) Plots of the I2930/I2850 ratio (red circles) measured
at pixels 1-5 along the lines a, b and c marked in (a). The noise level (black squares) calculated
at each pixel along each line is also plotted for comparison. The maximum change in the noise
level observed is up to 0.1 units, whereas the maximum change in the I2930/I2850 ratio is 2.4,
2.6 and 2.8 units along lines a, b and c, respectively. Since the change in I2930/I2850 signal is ca.
26× higher than the change in the noise level, we do not expect it to arise from instrumental
noise but from the heterogeneities in the lipid ordering at the nanoscale.

8
Figure S6. (a) Histogram of I2930/I2850 ratio in the TERS image shown in Figure 2f. Mean (µ) and
standard deviation (σ) of I2930/I2850 ratio in the TERS image are calculated as 2.14 and 0.77,
respectively. Regions of histogram within < µ - σ, µ ± σ, µ + σ to µ + 2.5σ and > µ + 2.5σ are
highlighted in blue, dark green, light green and yellow, respectively. (b) TERS image of the
I2930/I2850 ratio in which regions of intensity < µ - σ, µ ± σ, µ + σ to µ + 2.5σ and > µ + 2.5σ are
highlighted with the respective colors as shown in (a). In some regions of the image, I2930/I2850
ratio is > 2.5σ from the mean, which confirms the statistical significance of our TERS results.

9
Figure S7. (a) TERS image of the I2930/I2850 ratio of a patch of DPPC monolayer transferred onto
an Au (111) surface as shown in Figure 2f. (b) Plot of the TERS spectra measured at pixels 1-5
along the line marked in (a). The spectral intensity is normalized to the 2850 cm-1 peak. In the
spectra, a clear change in the relative intensity of 2930 cm-1 peak (with respect to the 2850
cm-1 peak) is observed indicating that the change in the I2930/I2850 ratio at the adjacent pixels
is due to heterogeneities in the lipid ordering at the nanoscale and not just due to the noise
of the measurement. (c) Mean and standard deviation of the peak at 2850 cm-1 and 2930 cm-
1 in the TERS spectra measured at pixels 1-5.

10
Figure S8. STM topography image of a DPPC monolayer transferred onto an Au (111) surface.
This region of the Au (111) surface contains several terraces and steps. TERS imaging of the
DPPC monolayer in this region is presented in Figure 4.

11
Figure S9. (a) TERS image of the I2930/I2850 ratio of a patch of DPPC monolayer transferred onto
an Au (111) surface as shown in Figure 4b. (b) Plot of the I2930/I2850 ratio (black squares) along
the pixels 1-5 along the line marked in (a). The noise level (red circles) calculated at each pixel
along the line is also plotted for comparison. The maximum change observed in the noise level
is 0.2 units, whereas the maximum change in the I2930/I2850 ratio is 0.6 units, which is 3× higher
than the change in noise level. This shows that change in I2930/I2850 ratio at the adjacent pixels
arises due to heterogeneities in the lipid ordering at the nanoscale and not the noise of the
measurement. (c) Plot of TERS spectra measured at pixels 1-5 along the line marked in (a).
Spectral intensity is normalized to the 2850 cm-1 peak. A clear variation in the relative intensity
of 2930 cm-1 peak is observed along the line as depicted in (b).

12
Figure S10. (a) TERS image of the I2930/I2850 ratio as shown in Figure 4b. (b) Plot showing the
mean and standard deviation of the 2850 cm-1 and 2930 cm-1 peaks for the TERS spectra
measured in regions 1a and 1b marked in (a). In region 1a, the intensity of the 2850 cm-1 and
2930 cm-1 peaks is 4.01 ± 0.45 and 3.83 ± 0.41, respectively. Whereas, in region 1b, it is 7.09
± 0.26 and 5.59 ± 0.20, respectively. The intensity of the 2850 cm-1 and 2930 cm-1 peaks is
comparable in region 1a, whereas, the intensity of the 2930 cm-1 peak is lower compared to
the 2850 cm-1 peak in region 1b signifying a relatively higher ordering of the lipid molecules
in region 1b. (c) Plot showing the mean and standard deviation of I2930/I2850 ratio at 1a (0.96 ±
0.03) and 1b (0.79 ± 0.03), respectively. A statistically significant difference in the mean value
is observed with a small standard deviation. Furthermore, the values are well separated
without any overlap indicating that the change in the I2930/I2850 ratio comes from the
difference in the relative ordering of the lipid molecules in the aforementioned regions, and
not the noise of the measurement. (d) TERS spectra measured in regions 1a (black) and 1b
(red). Spectral intensity is normalized to the 2850 cm-1 peak. A relative higher intensity of 2930
cm-1 peak is observed in region 1a as depicted in (c).

13
References

[1] J. Clavilier, R. Faure, G. Guinet, R. Durand, J. Electroanal. Chem. Interf. Electrochem. 1980, 107,
205-209.
[2] X. M. Yang, D. Xiao, S. J. Xiao, Y. Wei, Appl. Phys. A 1994, 59, 139-143.
[3] J. Stadler, T. Schmid, R. Zenobi, Nano Lett. 2010, 10, 4514-4520.
[4] H. Zhang, Q. Fan, Y. E. Wang, C. R. Neal, Y. Y. Zuo, Biochim Biophys Acta 2011, 1808, 1832-
1842.
[5] a) C. W. Hollars, R. C. Dunn, Biophys. J. 1998, 75, 342-353; b) V. Knecht, M. Müller, M. Bonn,
S.-J. Marrink, A. E. Mark, J. Chem. Phys. 2005, 122, 024704; c) M. Lösche, E. Sackmann, H.
Möhwald, Berichte der Bunsengesellschaft für physikalische Chemie 1983, 87, 848-852.
[6] J. Mikrut, P. Dutta, J. Ketterson, R. MacDonald, Phys. Rev. B 1993, 48, 14479.
[7] L. Chi, M. Anders, H. Fuchs, R. Johnston, H. Ringsdorf, Science 1993, 259, 213-216.
[8] a) F. Hubenthal, Appl. Phys. B 2014, 117, 1-5; b) A. Yaremko, V. Yukhymchuk, O. Hreshchuk, V.
Koroteev, V. Dzhagan, J. Raman Spectrosc. 2021, 52, 815-820; c) ^ M. Mori, S. Abe, T. Kondo,
Y. Saito, in Biomedical Vibrational Spectroscopy 2018: Advances in Research and Industry,
International Society for Optics and Photonics, 2018, p. 104900Z.
[9] a) E. Okamura, J. Umemura, T. Takenaka, J. Raman Spectrosc. 1991, 22, 759-762; b) R. G.
Snyder, S. L. Hsu, S. Krimm, Spectrochim. Acta A Mol. Spectrosc. 1978, 34, 395-406; c) R. C.
Spiker, I. W. Levin, Biochim. Biophys. Acta, Lipids Lipid Metab. 1975, 388, 361-373; d) C. Lee,
C. D. Bain, Biochim. Biophys. Acta, Biomembr. 2005, 1711, 59-71.

14