Professional Documents
Culture Documents
Wish Insights
3. Models
4. Result Interpretation
Business Understanding
online shopping platform
• Wish allows sellers to list products and sell directly to consumers. Wish does not
handle stock of product nor manage returns. Standard shipping that takes 2-3
weeks.
• 1M+ merchants offer cheap and unbranded goods on Wish platform. Wish’s
customers are typically working-class Americans who find Amazon Prime’s
membership fee is too expensive *
• 90M+ people use Wish at least once a month globally *
• Wish’s revenue was $1.9B in 2018 *
• Wish was valued at $11.2B in Aug 2019 series H funding round **
Price Provide low-priced no-name merchandise Higher price with control over quality of
from Chinese merchants product and customer service
MAXIMIZE SALES
Wish uses personalized Since almost all products are no- Attractive photo or price can get
shopping features, so everyone name merchandises, showing up consumers attention.
gets different product mix on in multiple pages can gain more
the screen. awareness of the product Wish marketplace use very
traditional screen design so
customer can only see a photo,
price, rating, and mark for rapid
shipment.
6
Forbes interview
SECTION 2
10 Columns Removed
Reason for Removal Columns Removed
Insufficient data differentiation Currency, theme
Removing
irrelevant data for No useful Insights (too many Title, product_url, product_picture, 1573, 30
unique categorical data points) product_id, merchant_title,
modelling
merchant_name,
merchant_info_subtitle,
merchant_has_profile_picture
8
Summary Statistics provide interesting insights
Data Mean/Median Range/Distributi
LOWER PRICES on
Great deals for consumers on Wish platform Units Sold 4339 Exponential
Distribution
A UNIQUE BUSINESS MODEL
Usually consumers must wait weeks for international Product Color Black 1 to 305
delivery
9
Data Preparation - 2
Removing Cleaning data by removing duplicate rows 1539,30
Duplicates
Dropping sizes
with few data 12 out of 69 unique sizes accounted for 94% of the data. So, dropping the remaining data 1450,38
points
11
Data Insight – 1
• Product-variation-inventory, countries-shipped-to,
inventory-total, and merchant-rating-count has
less influence on unit-sold. Thus, we decided to:
– drop countries-shipped-to and inventory-
total from data to reduce noise
– check product-variation-inventory and
merchant-rating-count in heatmap
Insights from
Pair Plots
4. SG has the largest product unit sold. We know that most products on Wish
are from China, so we were surprised from this finding. We checked the raw
data and found that only two items were shipped from SG. “XS-8XL Plus
Size Dresses Summer Fashion Clothes Womens Deep V-neck Sleeveless
Maxi Dresses Ladies Casual Spaghetti Strap Loose Beach Wear Floral
Printed Halter Party Dress” item sold 20,000pcs. Wish can utilize this
demand finding to generate more unit sales.
6. Too many color choices makes it hard for prediction. Wish can consider to
convert them into a smaller number of categories to improve prediction.
7. io
8. Urgency_text "Reduction sir less achils en gros’ affects units_sold. Wish can
combine the remaining two urgency texts into one.
Data prep based on Removing 11 columns such as user_ad_boosts, badge_local_produce as they have little to 1428,29
Insights from Plots no effect on the sales as seen in the graphs
18
SECTION 3
Models
Modeling Preparation – 1
Count
2 Type of Modeling Regression modeling Linear Regression, Ridge Regression, Lasso Regression,
Kneighbors Regressor, Polynomial Regression
Insights from
Regression
Modeling
On the heatmap:
– Unit sold has a significant
correlation with rating
count.
– XS size has a negative
correlation with product
variation inventory.
– Standard Scaler for KNeighbors is not a proper model, because it only has training set
score 0.33 and test set score 0.31.
– Polynomial Regression is not quite appropriate because most data are categorical.
Preparation for
1. We have divided into categories such that
Classification – there is balance of data points (rows) across each category
modeling – we are interested in understanding projected sales, anything
above 1000 for initial sale should be good and 1000 for binary
classification made more sense rather than 5000 or 10,000,
which are large quantities
– also there are fewer data points for values above 5000 so the
models will not be able to build good insight
2. It is more harmful to be
Insights from
Classification unable to predict the factor
Modeling that can generate unit sold
than finding a wrong positive,
so we used recall metrics.
Logistic Regression has a
recall result of 0.7 and is the
best among all models. So we
recommend Wish to use
Logistic Regression for
prediction.
Insights from
3. Comparing to ROC Curve and AUC for all Logistic
Classification Regression, Decision Tree and KNN Classifier, we found out
Modeling that Logistic Regression model has an AUC of 0.97, the best
among all. We confirmed that Logistic Regression model is
the best model to predict results.
Result Interpretation
What Model?
• Recommendation: • Recommendation:
Incentivize customers to leave Consider partnering with an existing
product reviews. seller to launch new products on Wish.
Next Steps
Next Steps
Son Nguyen
Nov 21, 2020
Table of Contents
I. Executive Summary ...........................................................................................2
II. Background .....................................................................................................2
i. Battery Costs.....................................................................................................2
ii. Vertical Integration ...........................................................................................3
III. Optionality .......................................................................................................3
IV. Lithium Market ..............................................................................................3
V. Lithium Mining Process .................................................................................5
VI. Valuation Model .............................................................................................6
i. Real Option Analysis ........................................................................................6
a. Underlying Asset ...........................................................................................7
b. Strike Price ....................................................................................................7
c. Time to Expiration .........................................................................................8
d. Risk-free Rate ................................................................................................8
e. Volatility ........................................................................................................8
ii. Real Option in Practice: Building Binomial Trees...........................................9
VII. Management’s Takeaways ...........................................................................12
Son Nguyen 1
I. Executive Summary
Tesla is considering investing in a lithium mine to decrease battery costs. This paper explores
how to appropriately value the rights to mine lithium in a 10,000-acre area in the Nevada desert.
The valuation method covered is a Real American call option. This model is superior to static
NPV analysis because of uncertainty in lithium prices and the mining technology necessary to
mine lithium affordably and safely. The paper includes background, the optionality, a discussion
of the lithium market and mining process, and an analysis of the valuation model.
II. Background
Tesla is an electric auto manufacturer who also produces myriad clean energy products. The goal
of the firm is to accelerate the world’s transition to zero emission vehicles [1].
In 2015, Tesla announced the creation of a new business segment: Energy Generation and
Storage. This segment initially produced home batteries and grid batteries. To support this
segment, the company built a massive factory deemed a Gigafactory in the Nevada Desert. This
factory has 5.3 million square feet of space. In 2020, the company announced it would start
producing its own battery cells. Later in 2020, Tesla vowed to cut battery costs by 50% to
provide a 25k electric vehicle [2].
i. Battery Costs
The way most technology works in our society today can be broken down using a smart phone.
The actual computing power of the smart phone takes up a fairly low amount of space. The
battery takes up a majority of available space. These batteries are all created using lithium. The
same principle goes for cars. Electric cars are mostly made up of batteries. Now, it stands to
reason, that if batteries comprise a large portion electric vehicles, that they also comprise a large
portion of the expenses on the income statement [3, 4].
As of 2019, Tesla’s batteries were roughly 20% of vehichle’s COGS. As battery technology has
taken drastic steps forward in recent years, the cost has come down from around $288 per kWh
in 2016 to roughly $130 in 2020 [5]. They still make up a large portion of Tesla’s vehicle costs.
Now, fast forward to February 2020, Tesla announced they would be launching their own battery
Son Nguyen 2
cell manufacturing plant in Fremont CA. Since they are venturing into their own battery creation
business, it makes sense to vertically integrate and acquire a means of lithium production too,
rather than purchasing it.
Historically, Tesla has relied almost completely on Panasonic to build its car batteries. Panasonic
was given a huge space in the Gigafactory to simplify logistics. Tesla announced they will begin
producing battery cells as of 2020. This vertically integrates their Tesla vehicle production from
assembly to part manufacturing. Vertically integrating even farther would be producing the raw
materials necessary to create vehicles [6]. To do this, Tesla is evaluating purchasing the mining
rights for a 10,000-acre area in the Nevada desert that is rich with lithium deposits. Should Tesla
purchase the right to mine lithium and source their own raw materials? If so, what are the mining
rights worth? To answer these questions, we will use an American call option with a small twist.
III. Optionality
With some background knowledge in hand, one can start framing this problem into the option
model. The optionality of this problem can be broken down as the following. First, the
underlying value of the American call option is the value of a fully operational lithium mine.
Said differently, it is the value of the cash flows harvested when a mine is opened and
successfully mines lithium. Second, there are two uncertainties in this valuation. One uncertainty
is the price of lithium in the future. The other uncertainty is the successful creation of the
necessary mining technology. The time to expiration on this option is 5 years. The purchased
mining rights would last for 10 years, but the 5-year option expiration allows an appropriate time
to build and facilitate the mining before the rights expire. The strike price is the cost of building
the lithium mine.
Son Nguyen 3
increased over the next few years, prices returned to pre-2016 levels. Despite being labeled as a
rare earth element, or REE, lithium is actually quite common. World lithium production
increased 300% between 2015 and 2018 [7]. While lithium demand continues to increase with
the expansion of the EV market, the future price of lithium is very uncertain. Tesla will refuse to
mine for lithium in Nevada if the current market price is below the cost of extraction. This cost
of extraction can be considered the strike price of our American call option.
10 Year High:
$23,500/ton
10 Year Low:
$4,500/ton
Son Nguyen 4
V. Lithium Mining Process
The other uncertainty affecting the option’s value is when the extraction technology will be
finalized. Tesla’s new mining process will extract lithium via sodium chloride from lithium rich
clay deposits. This process will likely involve pumping a hyper-saline solution into the ground
and forcing brine to the surface. Afterwards, the brine will likely be pumped into reservoirs and
left out to evaporate. Afterwards, the concentrated brine will be ready to be purified to separate
out the lithium. While this method is similar to existing mining techniques, it will be the first to
primarily use sodium chloride to extract the lithium found in Nevada’s clay deposits. These
deposits have not been mined previously since the lithium mining industry views clay deposits as
problematic. According to Whittaker, “sediments typically contain lower grades compared to
pegmatitic ‘hard rock’ minerals such as spodumene [8].” This could become an issue if the
lithium extracted by Tesla is unusable for battery cell production. Another negative associated
with these clay deposits is the lower recovery rate. A typical hard rock lithium mine will extract
95% or more of the lithium from the ground. Historical recovery rates for clay deposits average
around 35%. In order for Tesla to begin mining lithium, this technology must be completed
within the next five years and deliver a high enough recovery rate to be economical.
To incorporate this technical risk into our option’s valuation model, we estimated Tesla’s
probability of finishing the technology in each annual period. During each period we expect
Tesla to perform ten trials to test and perfect their mining technology. Each trial has a 5% chance
of the technology being successful. As each period passes, the likelihood of a scientific
breakthrough increases in the subsequent period. Once we have our first successful trial, we will
be able to open the mine if the lithium market’s conditions are favorable. For example, there is a
40% chance that Tesla will experience at least one successful trial in the first period.
Son Nguyen 5
VI. Valuation Model
Tesla’s management must value the mining rights and when to open the mine. The two risk
factors associated are uncertainty in future lithium prices (market risk) and uncertainty in mining
success (private risk). In other words, once they have the mining rights, do managers want to
invest immediately and take the mine, giving up the flexibility value of increasing lithium prices
and increasing technical success, or do they forego the early capture of the mine’s cash flows and
keep the flexibility?
At first glance, this seems like a valuable project and one that should be given the permission to
commence immediately. However, the static DCF model does not incorporate lithium price
volatility and the value to Tesla for waiting to open the mine. To incorporate the price volatility,
value for waiting and technical risk, we must analyze this project as a real option.
Figure 4: A hypothetical DCF model that does not capture technical risk and the value of waiting
Son Nguyen 6
Figure 5: Decision tree for opening/waiting to open the mine
If Tesla purchases the mining rights, they can initiate development and open the mine at any
time, as long as the technology works and benefits greater than costs. For simplicity reasons, we
assume that each year Tesla can develop the mine or wait. Thus, the resulting project can be
evaluated like an American call option illustrated by binomial trees.
a. Underlying Asset
The underlying asset represents the value of a fully operational lithium mine and production
facility. The asset valuation is closely similar to the traditional DCF model: the first component
is the expected cash flow harvested over the next year, which can be derived from next year’s
lithium prices (for simplicity we do not incorporate the net convenience yield in our
calculations); the second component is the expected value of the mine at the end of next year,
which is usually referred to as the “terminal value”. However, a key difference in our binomial
option model compared to the static DCF model is that there are two possible states (lithium
prices going up or down) once we move to the next year, so we need to sum the two components
in the “up” state and the “down” state, and then discount them by the associated risk-neutral
probabilities to get the underlying asset’s value at each point in our time horizon.
b. Strike Price
The strike price represents the total extraction cost needed for the mine to be operational. The
strike price is currently estimated at $7,000,000 in 2020 nominal dollars. It should be noted that
this extraction cost can rise in the future, but for simplicity we assume that at any time Tesla can
Son Nguyen 7
“exercise”, i.e., develop and open the mine by spending a fixed $7 million as total extraction
cost.
d. Risk-free Rate
Expecting a better economy after the pandemic and from 2021 onwards, we use a rather
generous 2% as the risk-free rate for this project, even though the current rate of the 10-year US
treasury bonds is below 1%. Changing the risk-free rate should not significantly change the
underlying asset’s and option’s values.
e. Volatility
Since historical data is limited, yet we know that the price of lithium has been fluctuated a lot
during 2018 and 2019, we can modestly assume that the volatility of lithium prices is around
30%. Regardless of volatility, the expected lithium price is the forward price under the risk-
neutral assumption. A high volatility would make the optionality more valuable. It should be
Son Nguyen 8
noted that this is the volatility for lithium prices only and not for the value of the underlying
asset, which can only be computed only after the pro forma cash flows are known.
The table below summarizes the parameters that will be used throughout our binomial trees.
The next table highlights key equations to compute the aforementioned parameters.
𝑒 (𝑟−𝑐)𝑡 − 𝐷
𝑈 = 𝑒 𝜎√∆𝑡 𝐷 = 1/𝑈 𝑞=
𝑈−𝐷
𝐷𝐸𝑅𝐼𝑉𝑈𝑃 = 𝑞 ∗ 𝑈𝑁𝐷𝑈𝑃 𝐷𝐸𝑅𝐼𝑉𝐷𝑁 = (1 − 𝑞) ∗ 𝑈𝑁𝐷𝐷𝑁
𝑞 ∗ 𝐷𝐸𝑅𝐼𝑉𝑈𝑃 + (1 − 𝑞) ∗ 𝐷𝐸𝑅𝐼𝑉𝐷𝑁
𝐷𝐸𝑅𝐼𝑉0 =
𝑒 𝑟∆𝑡
Figure 9: Equations for computing parameters used in binomial trees
The first step in our evaluation is to build the binomial tree for lithium prices. We take the initial
spot lithium price in November 2020 as $7 per kg, and multiply by U for the up state, and D for
the down state. We replicate this for every period of the tree, as can be seen below.
Son Nguyen 9
Figure 10: Binomial tree of lithium prices
Next, we need to figure out each year’s cash flow from an open, fully operational mine. Before
we do that, it is necessary to build the income statements for Tesla. At the most intuitive level, an
income statement would look like the one below, even though in real life it can be much more
complicated.
Figure 11: Income statement and free cash flow in one specific year, at one specific lithium price
The second step-building each year’s cash flow from a developed mine, is just to directly put the
free cash flows from the income statements into our binomial tree.
Son Nguyen 10
The third step is to calculate the value of an open mine at each associated lithium prices and
corresponding year. At each time point, we take the sum of the free cash flow over the next year
and the terminal value, repeat this for the up state and down state, and discount them one step
back by the risk neutral probabilities.
The fourth, final step is to build the binomial tree for the option on developing the lithium mine.
This is where we need to use a slight modification to account for technical risk. Specifically,
there is a probability of the mining technology being successful at each time point. On the one
hand, if Tesla wants to exercise the development option at each tollgate (they have five years to
do so), the payoff will be the value of an open mine (taken from the above binomial tree)
subtracted by the $7 million extraction cost, and then multiply the result by the probability of
success in the corresponding year. This would essentially reduce the value of exercising, because
if technology does not work then Tesla will not get anything from the mine. On the other hand,
the value of waiting is the next-period options values discounted by the risk-neutral probabilities.
Son Nguyen 11
Starting from the end and working backwards, we compare the value of exercising versus
waiting. If the former is higher, the option value is recorded as the value of exercising, and Tesla
should exercise the option by fully developing the mine, because on average this is more
valuable than waiting for the next period. Otherwise, the option value is recorded as the value of
waiting, and Tesla should postpone the development till later.
As a result, the American option model is the most appropriate model for our project, because as
time goes by, lithium prices may rise and there are more chance for successful mining, but value
of the mine shrinks rapidly. Tesla’s managers should take these into consideration to optimally
develop the mine and increase their retained earnings. Furthermore, it is important that if the
mining rights costs less than the option value at time 0 (2020), Tesla should purchase the rights
right now, because this is value creation.
Finally, since we are working with limited data available, a lot of factors can change in the
future. Managers can do sensitivity analysis to account for these changes. Some possible ways
are:
Son Nguyen 12
References
[1]
https://www.tesla.com/about#:~:text=Tesla's%20mission%20is%20to%20accelerate,to%20drive
%20than%20gasoline%20cars.
[2]
https://electrek.co/2020/02/11/tesla-building-pilot-battery-cell-manufacturing-line-fremont/
[3]
https://www.forbes.com/sites/greatspeculations/2020/01/13/how-battery-costs-impact-teslas-
margins-an-interactive-analysis/?sh=43cef1fe5036
[4]
www.forbes.com/sites/greatspeculations/2020/01/13/how-battery-costs-impact-teslas-margins-
an-interactive-analysis/?sh=43cef1fe5036.
[5]
https://www.investopedia.com/articles/investing/050715/economics-tesla-batteries.asp
[6]
https://corporatefinanceinstitute.com/resources/knowledge/strategy/vertical-integration/
[7]
https://www.usgs.gov/centers/nmic/lithium-statistics-and-information
[8]
https://www.linkedin.com/pulse/inside-teslas-lithium-clay-salt-extraction-process-alex-grant/
Son Nguyen 13
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
Son Nguyen
Joint with Dr. Seonguk Kim
Department of Mathematics
Depauw University
April 06, 2019
1 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
Contents
1. Introduction
1
2. The equation when f (x) = x
1
3. The equation when f (x) = x2
2 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
Introduction
1. Introduction
mu00 + ku = 0 (1)
u(0) = u(1) = 0 (D)
3 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
Introduction
4 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
Introduction
Question 1.
The eigenvalues of the equations are discrete like (1) even if the
function f (t) has a singular?
Question 2.
How about the smoothness of the corresponding eigenfunctions?
5 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
The equation when f (t) = 1
t
2.1 Discrete Specturum
1
2.The equation when f (t) = t
2.1 Discrete Specturum
6 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
The equation when f (t) = 1
t
2.1 Discrete Specturum
1.0
0.5
-0.5
-1.0
-1.5
From the values and graphs, we can see that the equation (3) has
discrete eigenvalues and its corresponding eigenvalues are smooth
at the origin even thought it has a singular weight.
8 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
1
3. The equation when f (t) = t2
3.1 Continuous Specturum
1
u00 (t) + λ u(t) = 0 (4)
t2
u(0) = u(1) = 0 (D)
9 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
Then u0 (t) = −t−1 v(s), u00 (t) = t−2 v 00 (s) + t−2 v 0 (s) and
10 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
3.1.1 λ < 0
√ √
−1+ 1−4λ −1− 1−4λ
α, 2 ,β , 2
⇒ C1 = 0, C2 = 0, ∴ v ≡ 0.
11 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
3.1.2 λ = 0
α = 0, β = −1
C1 + C2 = 0, C2 = 0 ∴ v ≡ 0.
12 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
1
3.1.3 0 < λ < 4
13 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
1
3.1.4 λ = 4
C is a arbitrary constant.
1
∴ u(t) = Ct 2 ln t.
14 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.1 Continuous Specturum
1
3.1.5 λ > 4
√
Since 1 − 4λ < 0
√ √ √ √
−1+ 1−4λ −1+ −ai −1− 1−4λ −1− −ai
α= 2 = 2 ,β = 2 = 2 ,
where a , 1 − 4λ
1
√ √
−a −a
⇒ General solution : v(s) = e− 2 (C1 cos 2 s + C2 sin 2 s)
15 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.2 Smootheness of Eigenfunctions at x = 0
λ ≤ 0 → u(t) = 0
1
0 < λ < → u(t) = t−α − t−β , α, β < 0
4
1 1
λ = → u(t) = t 2 ln t
4 √
1 − 12 −a
λ > → u(t) = t sin( ln t)
4 2
This shows that when it has a continous spectrum λ > 0 and each
corresponding solution has steep slope at the origin.
16 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
3.The equation when f (t) = 12
t
3.2 Smootheness of Eigenfunctions at x = 0
17 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
4. Conclusion and Future Works
18 / 19
Spectral Dynamics for the 2nd-Order Differential Equations with Singular Weight
4. Conclusion and Future Works
Thank you!!!
19 / 19
Zinc(II) Perchlorate Hexahydrate Catalyzed Opening of an Epoxide Ring by
Aniline Derivatives and Brine Shrimp Lethality Assay
Son Nguyen and Professor Jeffrey Hansen, Ph.D.
Department of Chemistry and Biochemistry, DePauw University, Greencastle, IN 46135
Winter 2019
3 2c 50% 28 ppm
dd
Figure 5. 13C NMR of N-Methylaniline product
Acknowledgements
I would like to thank the following people:
Add 1-2 drops of CH2Cl2 to the mixture if the amine - Professor Hansen for his constant
Fall 2018 4 2b 90% Not tested
reagent is a solid support and guidance throughout the
project
5 2d 81% 95 ppm*** - Professor Roberts for helping with
Stir the mixture at room temperature and monitor logistics and everything in the storeroom
progress by Thin Layer Chromatography (TLC) Unsuccessful - Marina Xie for her experience,
6 2e 79%
**** helpfulness and instructions
Winter 2019 - Noah Long for his previous work that set
Low yield up the foundation for this project
The reaction is done after 3-5 hours 7 2f
(< 3%)**
N/A
Figure 6. HSQC of N-Methylaniline product - Department of Chemistry and
Low yield
Biochemistry at DePauw University
8 2g N/A
(< 3%)**
Adsorb the product into silica gel and transfer to the top
of a chromatography column Figure 3. Results of the ring opening of our epoxide by aniline
References
derivatives and corresponding BSLA
*: Data in entries no. 1, 2 and 3 were retrieved from Noah Long’s poster. 1. Hansen, J.; Smith, Collin R.; Linder,
Purify by Flash Chromatography **: There were a lot of spots on the TLC plate when we analyzed the Ryan J.; Huffman, John C
reaction mixture, so we suspected that there were other compounds besides Diastereoselective sulfur ylide promoted
our main product that made purification really difficult. aldol/epoxidation, Tetrahedron Letters
Rotovap and vacuum to obtain pure product ***: This noticeably high data should be used with caution, because the 2006, 47 (40), 7209-7212 .
product was not soluble but rather formed precipitate readily with ASW when 2. Natongchai, W.; Khan, R.; Alsalme, A.;
we made the solution for BSLA. Shaikh, R. Catalysts 2017, 7(11), 340.
Record the mass, calculate % yield and identify product ****: Because of its poor solubility in ASW, even at the highest concentration Figure 7. gHMBC of N-Methylaniline product
by NMR the product was not able to kill 50% of the shrimp.