why where not in for nullable column gets different execution plan than column with not null

Question

I have a left table and a right table and both tables have 1 million records -

CREATE TABLE t_left (
id INT NOT NULL PRIMARY KEY,
value INT NOT NULL)

CREATE TABLE t_right (
id INT NOT NULL PRIMARY KEY,
value INT NOT NULL
)

sample data for t_left table is -

id  value
10000   1
20000   1
30000   1
40000   2
50000   2
60000   2

sample data for t_right table is -

id  value
2   2
3   2
4   2
5   2
6   2

Now I want to find all the rows of t_Left whose column "value" does not exists in column "Value" of table t_right

so my output should be

id  value
10000   1
20000   1
30000   1

Now my question is why sql server is generating different plan when column "Value" of t_left table is NOT NULL and very different plan when column "Value" is null. So as per above table definition my plan for query -

select a.id,a.value 
from dbo.t_left as a
where a.value not in (select b.value from t_right as b)

this is the plan when value column is not null

Now when change the column value from not null to null then I get total different plan. Can someone explain me this plan. Why the table t_right is scanned twice and why row count spool operator is showing in the plan. Which makes the query expensive. I need someone to help me out in understanding the plan in both the cases - when column is not null and when column is null.

alter table t_left
alter column value int null

I know I should be using where notexists when column is null. But why where not in is behaving differently in both the conditions