1. Broadband Microwave Spectrum & Structure of Cyclopropyl Cyanosilane
Nathan A. Seifert, Simon Lobsiger, Brooks H. Pate
University of Virginia
Gamil A. Guirgis, Jason S. Overby
College of Charleston
James R. Durig, Peter Groner
University of Missouri – Kansas City

2. Can we make CP-FTMW spectroscopy a valuable, but
Motivation
Can we make CP-FTMW spectroscopy a valuable, but
routine tool for synthetic chemists?
The immediate, but obvious requirement:
Sensitivity to consistently resolve structural information
But we also must consider:
Sample load; CP-FTMW is a destructive technique,
so we need to minimize sample consumption
Researcher – Spectrometer Interface: Fast, simple
tools to analyze spectra

3. Introduction
Last year, we reported microwave detection and structure
elucidation of a number of silicon-containing species (2013 RC12)
Spectra found using AUTOFIT with good results considering
complications from (for instance) internal rotation
And nitrogen-derived hyperfine splittings
Cyclopropyl cyanosilane:
Two conformers: cis- (shown) and gauche-
Again, all corresponding
rotational spectra were found using AUTOFIT, using
only the ab initio structure and its SPCAT predictions
as a guide

4. Dynamic range (gauche): 1300:1  Good enough for 15N!
Experimental
CP-FTMW, 6-18 GHz
5 nozzles, 250k averages
Dynamic range (cis): 6400:1
Dynamic range (gauche): 1300:1
 Good enough for 15N!
Sample load: ~1 mL liquid
vapor diluted to 0.2% in Ne
gauche conformer
Line densities:
~8000 lines ≥ 3:1 S:N
>3000 lines from 3:1 to 10:1
773 lines assigned

5. Beating Impurities Top AUTOFIT result:
Error window of ca. ±30 MHz
on isotopologue transitions (after scaling)
Manual assignment is possible,
but inefficient due to near-baseline
contaminants.
avg. 110 lines within 50%
predicted intensity per
transition search window
AUTOFIT to the rescue! ±30 MHz
window on fitting set of 3 transitions,
including 313 – 212:
~150k transition combinations
to check  10 minutes at 200 Hz
(standard for 4-6 CPU cores)
(gauche conformer)
Top AUTOFIT result:
Constants = , , ; OMC = 10.2 kHz
[Expt: (16), (17), (16)]

6. cis- conformer MP2/6-311++g(d,p) Parent 13C1 13C3 A /MHz 4291.0621
(84)
(28)
(62) B
(68)
(20)
(23) C
(60)
(11)
(13)
ΔJ / kHz
1.796
1.794(17)
1.799(44)
1.824(47)
ΔJK
-6.756
-6.295(29)
-6.04(20)
-6.03(18) ΔK
10.324
10.429(47)
10.33(38)
11.8(12) δJ
0.628
0.6035(37)
0.617(35)
0.653(37) δK
-0.473
-0.688(48)
[-0.688]
1.5(χaa) /MHz
-0.104
(88)
[ ]
0.25(χbb-χcc)
-1.194
(22)
[ ]
N lines
175 64 52
RMS fit (kHz)
12.5
14.6
12.8
13C4
29Si
30Si
15N
(61)
(20)
(21)
(93)
(18)
(14)
(13)
(29)
(12)
(10)
(95)
(18)
1.816(38)
1.784(32)
1.739(25)
1.695(40)
-6.14(18)
-6.17(12)
-6.17(15)
-6.65(16)
10.3(12)
9.97(23)
10.16(34)
11.0(18)
0.602(29)
0.620(16)
0.596(17)
0.525(38) -- 61 87 83 27
13.4
14.4
10.8

7. gauche- conformer MP2/6-311++g(d,p) Parent 13C1 13C2 13C3 A /MHz
(45)
(47)
(33)
(26) B
(48)
(48)
(38)
(36) C
(45)
(39)
(26)
(29)
ΔJ / kHz
0.745
0.866(24)
0.837(41)
0.759(30)
0.762(37)
ΔJK
-11.93(15)
-12.00(41)
-12.32(29)
-9.3(17) ΔK
69.006
73.39(93)
[73.39] δJ
0.212
0.2499(38)
0.258(45)
0.213(34)
0.195(33) δK
2.125
3.4(22)
[3.4]
1.5(χaa) /MHz
-4.037
-4.052(14)
-4.26(25)
-4.13(14)
-4.14(19)
0.25(χbb-χcc)
-0.564
(49)
-0.32(30)
-0.50(22)
-0.31(19)
N lines 80 15 19 14
RMS fit (kHz)
13.1
12.2
10.6
8.48
13C4
29Si
30Si
15N
(24)
(73)
(12)
(16)
(39)
(17)
(31)
(17)
(26)
(14)
(27)
(16)
0.794(27)
0.827(16)
0.846(24)
[0.866]
-12.94(32)
-12.40(24)
-12.41(24)
[-11.93]
0.229(34)
0.256(10)
0.272(31)
[0.2499]
-3.938(68)
-4.014(30)
-4.001(32) --
-0.578(17)
-0.567(18)
-0.518(31) 20 33 31 12
10.4
12.6
13.6
12.3

8. B3LYP-D3/aVTZ + ZPEharmonic
Relative Energies and Structural Statistics
Approximate expansion temperature measured at ~135 K in
hexanal data set [not a perfect comparison, but probably better than assuming STP!]
– see next talk!
Method
ΔE (cm-1)
σRMS, structure (Å)
Experimental
3 : 1 (~ K, K) --
MP2/ g(d,p)
203 (9 : 1) 135 K]
0.032 (gauche) / (cis)
MP2/ g(d,p) + ZPEharmonic
182 (87 : 13)
MP2/ g(d,p) [ ΔΔG135 ]
113 (77 : 23)
CCSD(T)/aVDZ
200 (9 : 1)
0.030 (gauche) / (cis)
B3LYP-D3/aVTZ
141 (82 : 18)
0.031 (gauche) / (cis)
B3LYP-D3/aVTZ + ZPEharmonic
120 (78 : 22)
Experimental ratio was least-squares fit by minimizing residuals
on “scale factor” between SPCAT and expt. intensities
σRMS is root mean square deviation between ab initio structure and Kraitchman determination

9. Gauche-Gauche Interconversion Tunneling?
Using a MP2/ g(d,p)
torsional potential:
V0 = 640 cm-1
Tunneling splitting
estimations:
via WKB approximation1,
ΔE01 = 304 kHz
via Pitzer’s
1D hindered rotor analysis2,
ΔE01 = 337 kHz
J. Bicerano, H. F. Schaefer III, W. H. Miller,
J. Am. Chem. Soc., 105, 2250 (1983).
2) K. S. Pitzer, W. D. Gwinn, J. Chem. Phys., 100
997 (1942).

10. Gauche-Gauche Interconversion Tunneling?
Doubled lineshape
seen on “high J” lines only
Not seen on cis- conformer
lines in this region, so likely
not inhomogenities
due to clustering
Problems:
CP-FTMW linewidth
too broad to make good
assignments beyond a
few lines with J ≥ 5
Transitions across the
gap (μc-type) predicted
to be very weak,
with μc ≤ 0.2 D (for
comparison, μa ~ 4 D)

11. New Software in Development
CALPGM API
A Python wrapper for SPCAT and SPFIT made for fast and easy access to
fitting and prediction routines
Stores predictions and fits in memory, with the ability to dynamically change floated/fixed
parameters as needed
Automatic parsing/generation of input and output files
Can search and filter SPCAT predictions quickly via frequency, intensity, and quantum
numbers
Interfaces with Numpy for fast analysis and integration into future user interfaces
for rotational spectrum analysis (e.g. AABS) – spectral predictions for
asymmetric tops can update smoothly (e.g. with JB95-style constant scroll bars)!
Potentially reimplementing JB95-type linearization for faster predictions

12. New Software in Development
AUTOFIT “2.0”
Complete rewrite of the original AUTOFIT code for better conciseness and optimization
Using the CALPGM API, early benchmarking suggests at least a 40% speed improvement
Redone and streamlined user interface using industry standard JSON input data formatting
Better fit refinement tools --- automated eQq / distortion constant assignment?
More options for defining search parameters and constraints
Dynamically allocated and non-symmetric search windows
Absolute/relative intensity constraints
Automatic determination of triples with Ka/Kc / relative intensity constraints
Potential integration with cloud services such as Amazon EC2 for extremely fast, parallelized
AUTOFIT searches (≥ 3000 Hz?)

13. Questions? Thank you for your time!
This work was funded by the National Science Foundation’s Major Research Instrumentation program, award # CHE