1. California Institute of Technology
An Asynchronous Optical Sampling THz-TDS Spectrometer and THz Frequency Comb Spectrometer
Jacob Good, Dan Holland, Ian Finneran, Brandon Carroll, Marco Allodi, Geoff Blake
California Institute of Technology

2. Gas-Phase Rotational Spectroscopy
D2O
10 THz/100 kHz
10-8 precision
10 ppb
Gas-Phase rotational spectroscopy of small molecules overlaps with the bulk of the THz region from THz.
In order to assign transitions using quantum theory, we need to measure transitions to a precision of 10 ppb in order to compare with theory.

3. What does 10-8 Precision Look Like?
In terms of length, measuring the diameter of a tennis ball in units of nm (width of a glucose molecule) would represent 10 ppb precision.
Tunable CW sources have been able to achieve greater than 1 ppb precision.
However, they require expensive electronics and require a prohibitively long time to scan across the entire THz region.
108 nm to 1 nm
10-8 precision
10 ppb
publications.nigms.nih.gov/chemhealth/images/ch4_size.jpg

4. Part I: Building the Spectrometer
Part I answers the question:
Why do we need a unique spectrometer for this work?
Part I: Building the Spectrometer

5. Time Domain Spectroscopy
Fortunately, TDS has already developed a methodology for measuring units of time with extreme precision.
An ultrafast laser emits extremely short pulses of light at an extremely precise interval.
These pulses can be used to initiate an ultrafast process of interest, but the electronics to directly measure this process do not exist.
If the ultrafast process is repetitive, we can circumvent this problem by splitting off a small amount of light from the laser and sending it to a delay stage to add a variable amount of delay.
We can use these variably delayed pulses to sample the process of interest across the entire time interval, effectively slowing down time.
coherent.com;
Dan Holland, Caltech PhD Thesis, (2014)

6. Delay Line Scan Rate vs Scan Range
Unfortunately, even obtaining 1 ppm precision is too demanding for a mechanical delay stage.
More delay time requires a longer stage and precision mechanical motors can only move so quickly.
To get to 1 ppm we need 100 ns of delay which would require a 30 m long delay stage, longer than most laser tables!
10 THz to 10 GHz
10-3 precision
1 ppm
10 THz
10 GHz
Dan Holland, Caltech PhD Thesis, (2014)

7. ASynchronous OPtical Sampling
Master
Slave
ASOPS is a method that uses a separate laser to emit a delayed probe pulse train, eliminating the need for a delay stage entirely.
ASOPS requires extremely precise asynchronization of two lasers but with no moving parts ASOPS is far more precise than a delay stage based setup.
coherent.com;
Dan Holland, Caltech PhD Thesis, (2014)

8. ASOPS Sweep Rate vs Scan Range
80 THz
With ASOPS, we can generate an arbitrarily long delay by choosing the repetition rate of our master laser.
We choose our high-frequency resolution and scan speed by choosing the repetition rate offset between master and slave lasers.
We can obtain plenty of high-frequency resolution and scan speed, however we will need to use data processing methods to improve our precision.
10 THz to 10 MHz
10-6 precision
1 ppm
10 MHz
Dan Holland, Caltech PhD Thesis, (2014)

9. THz Electro-Optic Detection
ASOPS-THz-TDS System
THz Electro-Optic Detection
Balanced
PDs
Personal Computer
180 MS/s Digitizer
Wollaston Prism
λ/4
ZnTe
10 MHz Rb Standard
Lens
Periscope
Master
Ti:Sapph
ITO
Slow PZT
Scan Triggering
Sample
Cell
12.5 cm
Frequency Counter
50%
90%
Clock PD
PMT
BBO
Slave
Ti:Sapph
Periscope
Slave PD
The electronic circuit that we just built doesn’t actually keep track of when the pulse trains of the Master and Slave lasers synchronize.
In order to know when to start a new scan, we use a cross-correlator that generates pulses of blue light and converts them to an electronic trigger signal.
Since we want to do THz spectroscopy we use a THz emitter to convert our pulses of optical light into pulses of THz light.
Water vapor in the air strongly absorbs THz photons so we need to house the THz portion of our beam in a purge box.
We route our THz beam through our sample cell and onto a ZnTe crystal. ZnTe changes its birefringence in response to THz pulses.
We then route our optical probe beam onto this crystal where its polarization changes in response to the THz induced birefringence changes.
We then extract the component polarizations of the beam and measure the magnitude of the polarization change on a pair of balanced photodiodes.
We precisely digitize the resulting signal and can then use FFT data processing to extract its frequency content.
THz
Emitter
Slow PZT
Fast PZT
50%
Lens
Master PD
PLL Electronics
50%
Nitrogen Purge Box
90%

10. 80 MHz ASOPS THz Pulse J.T. Good, et al., Rev. Sci. Instrum., In Prep
We now have the raw data that we need to extract frequency content. This data is sampled in the time domain, so the rest of the talk will discuss
Data processing to produce high resolution gas-phase spectra.
J.T. Good, et al., Rev. Sci. Instrum., In Prep

11. 80 MHz ASOPS Bandwidth J.T. Good, et al., Rev. Sci. Instrum., In Prep
1 hour integration
J.T. Good, et al., Rev. Sci. Instrum., In Prep

12. Part II: Gas-Phase Spectroscopy

13. Water 1 Torr − 80 MHz ASOPS
1 hour integration
J.T. Good, et al., Rev. Sci. Instrum., In Prep

14. Pressure Broadened Line Shape
-2.67 MHz
Observed - Calculated
114 MHz FWHM
Lorentzian Line Shape
10 THz to 10 MHz
10-6 precision

15. CH3CN 2 Torr − 80 MHz ASOPS ν = 2B(J + 1) − 2DJKK2(J + 1)
1 hour integration
J.T. Good, et al., Rev. Sci. Instrum., In Prep

16. CH3CN − Unresolved Structure
ν = 2B(J + 1) − 2DJKK2(J + 1)

17. 5 MHz FCS THz Pulse Train
J.T. Good, et al., Rev. Sci. Instrum., In Prep

18. 18 Pulse Frequency Comb
I.A. Finneran, J.T. Good, et al., Phys. Rev. Lett. 114, , (2015)

19. 5 MHz FCS Bandwidth J.T. Good, et al., Rev. Sci. Instrum., In Prep
1 hour integration
J.T. Good, et al., Rev. Sci. Instrum., In Prep

20. CH3CN 150 mTorr − 5 MHz FCS
1 hour integration
J.T. Good, et al., Rev. Sci. Instrum., In Prep

21. 5 MHz FCS vs 80 MHz ASOPS 3 J = 46 K = 6 2 1 4 5
J.T. Good, et al., Rev. Sci. Instrum., In Prep

22. Doppler Broadened Line Shape
0.086 MHz
Observed - Calculated
4.7 MHz FWHM
Gaussian Line Shape
10 THz to 100 kHz
10-8 precision

23. MeOH 140 mTorr FCS

24. MeOH 140 mTorr FCS

25. MeOH 140 mTorr FCS

26. MeOH 140 mTorr Peak Fitting
8.5 MHz FWHM
Gaussian Line Shape

27. MeOH 140 mTorr Prediction (THz) Observed (THz) Log Intensity
Error (MHz)
0.730
Fixed
N/A
-0.532
-0.775
-0.184
1.668
-0.031
RMS 0.54 MHz

28. THz Electro-Optic Detection
Long Path Double Pass Cell
THz Electro-Optic Detection
Balanced
PDs
Personal Computer
180 MS/s Digitizer
Wollaston Prism
λ/4
ZnTe
10 MHz Rb Standard
Lens
Periscope
Master
Ti:Sapph
ITO
Slow PZT
Scan Triggering
Frequency Counter
50%
90%
Clock PD
PMT
BBO
Slave
Ti:Sapph
Periscope
Slave PD
THz
Emitter
Slow PZT
Fast PZT
1 hour integration
50%
Lens
Master PD
PLL Electronics
50%
90%
Sample Cell 2x 0.94 m
188 cm / 12.5 cm = 15x enhancement
THz
Polarizer
Rooftop
Reflector

29. 21,470 Pulse Frequency Comb 10 THz to 10 kHz >10-9 precision
I.A. Finneran, J.T. Good, et al., Phys. Rev. Lett. 114, , (2015)

30. THz Bandwidth vs Pulse Width
P.J. Hale, et al., Opt. Express, 22, , (2014)

31. Prism Compressor

32. Acknowledgements
Blake Group
Prof. Jun Ye
NSF Funding

33. THF 5 Torr − 80 MHz ASOPS
1 hour integration

34. THF − Pseudorotational ModesQ Q P R P R P R P R